# Cobalt

Insights for security leaders, pentesters and developers -- all in one place. Learn more about how the world of App- and InfoSec is changing every day.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## 5 Takeaways From the Autonomous Pentesting Pulse Report: Where AI Fits in Continuous Pentesting

DevFeed: [5 Takeaways From the Autonomous Pentesting Pulse Report: Where AI Fits in Continuous Pentesting](<https://devfeed.tech/articles/5-takeaways-from-the-autonomous-pentesting-pulse-report-where-ai-fits-in-continuous-pentesting-137870.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/5-takeaways-from-the-autonomous-pentesting-pulse-report>)

Author: Gunter Ollmann

Published: 2026-11-17T14:53:48Z

Content type: news

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [human review](<https://devfeed.tech/topics/human-review.md>), [trust](<https://devfeed.tech/topics/trust.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [attack-paths](<https://devfeed.tech/tags/attack-paths.md>), [attack-surface](<https://devfeed.tech/tags/attack-surface.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-agents](<https://devfeed.tech/tags/autonomous-agents.md>), [autonomous-pentesting](<https://devfeed.tech/tags/autonomous-pentesting.md>), [continuous-pentesting](<https://devfeed.tech/tags/continuous-pentesting.md>), [pentesting](<https://devfeed.tech/tags/pentesting.md>), [pulse](<https://devfeed.tech/tags/pulse.md>), [report](<https://devfeed.tech/tags/report.md>), [research](<https://devfeed.tech/tags/research.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>)

### AI overview

The report's surveys suggest AI use in pentesting is growing, while trust in fully autonomous testing remains limited. Among Cobalt Core pentesters, frequent or extensive AI use rose from 32% in January 2026 to 61% in September. AI's leading reported use shifted from report writing to brainstorming attack paths, and 59% said exploit development and proof-of-concept work had shifted significantly to AI. Autonomous testing was rated more favorably over time, but 94% still ranked a human-led model ahead for complex, high-impact vulnerabilities; only 2% would let AI chain an exploit into a multi-step attack without supervision. The report recommends combining autonomous testing for breadth and frequent coverage with human expertise for business logic, attack chains, and validation, and adjusting that balance as capability and cost change. In a separate Omdia survey of 400 North American security professionals, 94% expected agentic AI to become critical or very important within two years, while 7% reported using it at full autonomy today. The article notes reliability and compute cost as constraints.

### Source excerpt

Every security leader I talk to is hearing two messages about AI at once. Autonomous agents will soon find and exploit vulnerabilities faster and more cheaply than people can. And you shouldn't trust them to do it alone yet. Both messages are partly true, and the decisions you make about testing over the next two years depend on getting the balance right.

## The Hidden Cost of Building Your Own AI Pentest, From a Four-Time CISO

DevFeed: [The Hidden Cost of Building Your Own AI Pentest, From a Four-Time CISO](<https://devfeed.tech/articles/the-hidden-cost-of-building-your-own-ai-pentest-from-a-four-time-ciso-129368.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/the-hidden-cost-of-building-your-own-ai-pentest-from-a-four-time-ciso>)

Author: Anne Nielsen

Published: 2026-11-10T19:17:54Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>)

Tags: [2015](<https://devfeed.tech/tags/2015.md>), [accountability](<https://devfeed.tech/tags/accountability.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-accountability](<https://devfeed.tech/tags/ai-accountability.md>), [ai-pentesting](<https://devfeed.tech/tags/ai-pentesting.md>), [ai-tool](<https://devfeed.tech/tags/ai-tool.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [attack](<https://devfeed.tech/tags/attack.md>), [automate](<https://devfeed.tech/tags/automate.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-pentesting](<https://devfeed.tech/tags/autonomous-pentesting.md>), [build](<https://devfeed.tech/tags/build.md>), [ciso](<https://devfeed.tech/tags/ciso.md>), [cost](<https://devfeed.tech/tags/cost.md>), [human-in-the-loop](<https://devfeed.tech/tags/human-in-the-loop.md>), [pentest](<https://devfeed.tech/tags/pentest.md>), [pentesting](<https://devfeed.tech/tags/pentesting.md>), [question](<https://devfeed.tech/tags/question.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>), [time](<https://devfeed.tech/tags/time.md>), [tony](<https://devfeed.tech/tags/tony.md>), [tool](<https://devfeed.tech/tags/tool.md>)

### AI overview

Building an AI pentesting tool in-house may be a useful side project or red-team aid, but the article argues against relying on it for enterprise testing. The organization assumes responsibility when something breaks, lacks independent review and risk transfer, and must keep the tool current and assess whether its findings remain reliable. Free or open-source inputs do not remove the ongoing labor or supply the pentesting experience needed for trustworthy results. The piece recommends external testing for that assurance, citing Cobalt's 13 years of pentesting, established methodology, continuous approach, and human direction as features the author says Tony Spinelli would trust. It does not claim experimentation has no value; it cautions against treating a homegrown tool as an accountable enterprise control.

### Source excerpt

I recently sat down with Tony Spinelli for a fireside chat on AI, accountability, and what still needs a human in the loop. But one audience question stuck with me, because almost everyone has thought it at some point: why can't I just build an autonomous pentesting tool myself?

## Cobalt + Wiz: Bringing Real-World Offensive Security Into Your Cloud Risk Workflow

DevFeed: [Cobalt + Wiz: Bringing Real-World Offensive Security Into Your Cloud Risk Workflow](<https://devfeed.tech/articles/cobalt-wiz-bringing-real-world-offensive-security-into-your-cloud-risk-workflow-119366.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/cobalt-wiz-bringing-real-world-offensive-security-into-your-cloud-risk-workflow>)

Author: Brittany Walker

Published: 2026-11-05T16:13:34Z

Content type: release

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [cloud security](<https://devfeed.tech/topics/cloud-security.md>), [Cloud & Application Security](<https://devfeed.tech/topics/cloud-application-security.md>), [Security](<https://devfeed.tech/topics/security.md>)

Tags: [account](<https://devfeed.tech/tags/account.md>), [act](<https://devfeed.tech/tags/act.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-application](<https://devfeed.tech/tags/ai-application.md>), [all](<https://devfeed.tech/tags/all.md>), [anchor](<https://devfeed.tech/tags/anchor.md>), [application](<https://devfeed.tech/tags/application.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-security](<https://devfeed.tech/tags/cloud-security.md>), [exposed](<https://devfeed.tech/tags/exposed.md>), [gap](<https://devfeed.tech/tags/gap.md>), [great](<https://devfeed.tech/tags/great.md>), [news](<https://devfeed.tech/tags/news.md>), [offensive](<https://devfeed.tech/tags/offensive.md>), [offensive-security](<https://devfeed.tech/tags/offensive-security.md>), [platforms](<https://devfeed.tech/tags/platforms.md>), [risk](<https://devfeed.tech/tags/risk.md>), [security](<https://devfeed.tech/tags/security.md>), [that-s](<https://devfeed.tech/tags/that-s.md>), [with](<https://devfeed.tech/tags/with.md>), [wiz](<https://devfeed.tech/tags/wiz.md>)

### AI overview

The Cobalt-Wiz integration brings pentest evidence into Wiz to help teams distinguish exposed issues from risks demonstrated as exploitable. Wiz provides visibility into vulnerabilities, misconfigurations, and potential attack paths; Cobalt adds findings from human pentesters, including evidence of how issues can be chained into an attack. Findings are sent automatically to Wiz's Pentest Table with asset and target details, severity, exploitability evidence, and remediation guidance, and stay synchronized with Cobalt. The integration is generally available through the Wiz Integration Network to Cobalt customers with integration access. Setup uses a Wiz service account with scoped permissions; the text says no additional Cobalt license is required.

### Source excerpt

Cloud security platforms are great at telling you what's exposed. They're not built to tell you what an attacker could actually do with it. That's the gap Cobalt and Wiz are closing together.

## AI's Biggest Security Risks Are Increasingly Familiar Flaws

DevFeed: [AI's Biggest Security Risks Are Increasingly Familiar Flaws](<https://devfeed.tech/articles/ai-s-biggest-security-risks-are-increasingly-familiar-flaws-114863.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/ai-biggest-security-risks>)

Author: john.zorabedian@cobalt.io (John Zorabedian)

Published: 2026-11-02T19:29:57Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Security research](<https://devfeed.tech/topics/security-research.md>), [Memory Safety](<https://devfeed.tech/topics/memory-safety.md>), [AI software supply chain security](<https://devfeed.tech/topics/ai-software-supply-chain-security.md>), [prompt injection](<https://devfeed.tech/topics/prompt-injection.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [access-controls](<https://devfeed.tech/tags/access-controls.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-chips](<https://devfeed.tech/tags/ai-chips.md>), [ai-risk](<https://devfeed.tech/tags/ai-risk.md>), [ai-security](<https://devfeed.tech/tags/ai-security.md>), [ai-security-risks](<https://devfeed.tech/tags/ai-security-risks.md>), [ai-security-testing](<https://devfeed.tech/tags/ai-security-testing.md>), [ai-stack](<https://devfeed.tech/tags/ai-stack.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [apis](<https://devfeed.tech/tags/apis.md>), [applications](<https://devfeed.tech/tags/applications.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [axera](<https://devfeed.tech/tags/axera.md>), [buffer-overflow](<https://devfeed.tech/tags/buffer-overflow.md>), [c](<https://devfeed.tech/tags/c.md>), [c-plus-plus](<https://devfeed.tech/tags/c-plus-plus.md>), [cameras](<https://devfeed.tech/tags/cameras.md>), [prompt-injection](<https://devfeed.tech/tags/prompt-injection.md>), [research](<https://devfeed.tech/tags/research.md>), [risks](<https://devfeed.tech/tags/risks.md>), [security](<https://devfeed.tech/tags/security.md>)

### AI overview

Cobalt reports that familiar software flaws are a growing part of AI security findings, alongside serious issues that often remain unresolved. In its AI and LLM application pentests, the LLM-specific share of findings fell from 84.5% in 2025 to 59.9% through September 2026, while conventional and unclassified findings grew 320% by volume. Only 38.4% of high-risk findings were resolved. Separately, an audit of the ax-llm runtime found seven defects, including memory-safety flaws and OS command injection; upstream fixes were shipped, but they reach devices only when vendors update firmware and owners install it. Cobalt recommends testing applications, APIs, infrastructure, access controls and configurations, tracking high-risk resolution rates and time to fix, and requiring vendor testing evidence and confirmation that fixes are present in deployed versions. Its pentest statistics concern AI applications, not edge hardware or firmware.

### Source excerpt

When security teams talk about AI risk, the conversation usually starts with prompt injection, jailbreaks, and guardrail evasion. Those threats are real. But the newest data from Cobalt pentests points somewhere less exotic: the flaws showing up most in AI systems are increasingly the same ones security teams have been fighting for decades, and most of the serious ones are still open.

## Supabase From a Pentester's Perspective: Methodology and Where the Real Risks Hide

DevFeed: [Supabase From a Pentester's Perspective: Methodology and Where the Real Risks Hide](<https://devfeed.tech/articles/supabase-from-a-pentester-s-perspective-methodology-and-where-the-real-risks-hide-109761.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/supabase-from-a-pentesters-perspective-methodology-and-where-the-real-risks-hide>)

Author: Lukasz Wierzbicki

Published: 2026-10-30T17:39:23Z

Content type: tutorial

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Supabase](<https://devfeed.tech/topics/supabase.md>), [code security](<https://devfeed.tech/topics/code-security.md>), [LocalStorage](<https://devfeed.tech/topics/localstorage.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [access-tokens](<https://devfeed.tech/tags/access-tokens.md>), [api](<https://devfeed.tech/tags/api.md>), [api-access](<https://devfeed.tech/tags/api-access.md>), [api-keys](<https://devfeed.tech/tags/api-keys.md>), [app](<https://devfeed.tech/tags/app.md>), [application](<https://devfeed.tech/tags/application.md>), [applications](<https://devfeed.tech/tags/applications.md>), [attack-surface](<https://devfeed.tech/tags/attack-surface.md>), [auth](<https://devfeed.tech/tags/auth.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [automation](<https://devfeed.tech/tags/automation.md>), [backend](<https://devfeed.tech/tags/backend.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [cobalt-core](<https://devfeed.tech/tags/cobalt-core.md>), [methodology](<https://devfeed.tech/tags/methodology.md>), [pentester](<https://devfeed.tech/tags/pentester.md>), [perspective](<https://devfeed.tech/tags/perspective.md>), [risks](<https://devfeed.tech/tags/risks.md>), [supabase](<https://devfeed.tech/tags/supabase.md>)

### AI overview

Supabase's public client key is expected; the real risk is weak authorization, missing or overly permissive Row Level Security (RLS), exposed privileged keys, or unsafe storage settings. The pentest method starts by finding project configuration and service endpoints in client code and traffic, then probing REST, Auth, Storage, functions, and other enabled services. Test known records across unauthenticated and authenticated users, vary queries and identifiers, and check reads and writes across roles to build an authorization matrix. Extend checks to storage, Realtime, functions, and common web risks such as rate limits, token storage, and privileged function logic. Defenders should keep service_role keys server-side, test access rules across resources and roles, review storage exposure, and correlate platform logs for broad queries or unusual access patterns.

### Source excerpt

Methodology and where the real risks hide.

## Top 10 Spookiest Web App Vulnerabilities

DevFeed: [Top 10 Spookiest Web App Vulnerabilities](<https://devfeed.tech/articles/top-10-spookiest-web-app-vulnerabilities-109762.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/top-10-spookiest-web-app-vulnerabilities>)

Author: Kyle Goode

Published: 2026-10-30T16:46:53Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [code security](<https://devfeed.tech/topics/code-security.md>), [Web app](<https://devfeed.tech/topics/webapp.md>), [XSS](<https://devfeed.tech/topics/xss.md>), [Security research](<https://devfeed.tech/topics/security-research.md>), [Adversarial attacks](<https://devfeed.tech/topics/adversarial-attacks.md>), [software security audit](<https://devfeed.tech/topics/software-security-audit.md>), [Static code analysis](<https://devfeed.tech/topics/static-code-analysis.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [appsec](<https://devfeed.tech/tags/appsec.md>), [research](<https://devfeed.tech/tags/research.md>), [vulnerabilities](<https://devfeed.tech/tags/vulnerabilities.md>), [web-app](<https://devfeed.tech/tags/web-app.md>)

### AI overview

Missing access control dominates the supplied 2026 web app pentest data at 41.9% of findings; together with cross-site scripting and server-side injection, it accounts for 75.2% of grave findings. The text attributes access control's persistence to scanners' inability to judge intended permissions, many authorization checks across APIs and objects, and rapid feature development. It recommends explicit authorization testing across roles, accounts, and tenants, alongside prioritizing output encoding and input handling. Authentication and session findings rose even as overall findings fell, while XSS declined but remained common. The text also reports that AI/LLM environments had a higher share of high or critical findings than traditional software and calls for strict tenant isolation and limits on agent actions. Business logic and access control flaws require human testing beyond clean automated scans.

### Source excerpt

Vulnerabilities don't need hockey masks or dirty jackets to frighten people. Whether AppSec knows they exist or not, these threats sit in the shadows, waiting to be weaponized.

## Pentester Spotlight--Prateek Thakare

DevFeed: [Pentester Spotlight--Prateek Thakare](<https://devfeed.tech/articles/pentester-spotlight-prateek-thakare-105918.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/pentester-spotlight-prateek-thakare>)

Author: Noelle Hori

Published: 2026-10-28T19:26:25Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Code review](<https://devfeed.tech/topics/code-review.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [REST API](<https://devfeed.tech/topics/rest-api.md>), [Web](<https://devfeed.tech/topics/web.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [ai](<https://devfeed.tech/tags/ai.md>), [api](<https://devfeed.tech/tags/api.md>), [apis](<https://devfeed.tech/tags/apis.md>), [bug-bounty](<https://devfeed.tech/tags/bug-bounty.md>), [burp-suite](<https://devfeed.tech/tags/burp-suite.md>), [business-logic](<https://devfeed.tech/tags/business-logic.md>), [cobalt-core](<https://devfeed.tech/tags/cobalt-core.md>), [code-review](<https://devfeed.tech/tags/code-review.md>), [communication](<https://devfeed.tech/tags/communication.md>), [defense-in-depth](<https://devfeed.tech/tags/defense-in-depth.md>), [interview](<https://devfeed.tech/tags/interview.md>), [pentester](<https://devfeed.tech/tags/pentester.md>), [spotlight](<https://devfeed.tech/tags/spotlight.md>)

### AI overview

Prateek Thakare says his work centers on finding flaws in business logic and access controls, informed by code review and by learning how an application is meant to work. He uses Burp Suite and extensions for manual testing and repetitive tasks, Waymore to find archived endpoints and files, and MCP servers for supporting tasks. He also tests web applications, APIs, and mobile apps. To reduce missed issues, he maintains a checklist, takes thorough notes, and keeps learning. His reports use screenshots, step-by-step reproduction instructions, and enough application context for readers with different levels of security knowledge. He argues that pentesters should help customers fix vulnerabilities and that security should be built into the development lifecycle. He expects AI to accelerate testing but sees strong results as depending on experienced testers' judgment, alongside AI tools. He also calls layered defenses a baseline as AI raises capabilities for attackers and defenders.

### Source excerpt

The Cobalt Pentester Spotlight highlights the fascinating journey of our Core members. Through an interview style, we share their experiences, background, and insights into the world of an accomplished pentester.

## The Patch Is the Starting Gun: PHP2Shell and the AI Race to a PoC

DevFeed: [The Patch Is the Starting Gun: PHP2Shell and the AI Race to a PoC](<https://devfeed.tech/articles/the-patch-is-the-starting-gun-php2shell-and-the-ai-race-to-a-poc-103663.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/the-patch-is-the-starting-gun-php2shell-and-the-ai-race-to-a-poc>)

Author: Joe Brinkley

Published: 2026-10-27T17:37:16Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Threat monitoring](<https://devfeed.tech/topics/threat-monitoring.md>), [Linux](<https://devfeed.tech/topics/linux.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [arm64](<https://devfeed.tech/tags/arm64.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [aslr](<https://devfeed.tech/tags/aslr.md>), [attack-surface](<https://devfeed.tech/tags/attack-surface.md>), [audit](<https://devfeed.tech/tags/audit.md>), [bugs](<https://devfeed.tech/tags/bugs.md>), [calif](<https://devfeed.tech/tags/calif.md>), [campaign](<https://devfeed.tech/tags/campaign.md>), [code-execution](<https://devfeed.tech/tags/code-execution.md>), [critical](<https://devfeed.tech/tags/critical.md>), [cve](<https://devfeed.tech/tags/cve.md>), [cve-2026-109618](<https://devfeed.tech/tags/cve-2026-109618.md>), [cve-2026-109619](<https://devfeed.tech/tags/cve-2026-109619.md>), [october](<https://devfeed.tech/tags/october.md>), [patch](<https://devfeed.tech/tags/patch.md>), [patches](<https://devfeed.tech/tags/patches.md>), [php](<https://devfeed.tech/tags/php.md>), [php2shell](<https://devfeed.tech/tags/php2shell.md>), [poc](<https://devfeed.tech/tags/poc.md>), [remote-code-execution](<https://devfeed.tech/tags/remote-code-execution.md>), [research](<https://devfeed.tech/tags/research.md>)

### AI overview

Attackers reportedly began exploiting PHP2Shell four days before patches shipped on October 21, and AI can speed up turning patch differences into candidate PoCs. Public PoCs are not proof of exposure: many reportedly fail on common configurations, including systems with ASLR enabled, and a crash only shows that a script failed on that particular setup. Check vendor advisories because Linux distributions may backport fixes without changing version strings. Watch for bursts of similar, slightly varied requests and PHP worker crashes or restarts; VulnCheck released Suricata rules for the traffic. Inventory exposed PHP, including forgotten apps and embedded appliances, prioritize unsupported PHP 8.1 and older, and verify fixes with tests against your own systems. ASLR value 2 raises attack cost but does not fix the flaw. Track temporary mitigations with owners and expiry dates until the real fix is deployed.

### Source excerpt

PHP2Shell is two critical remote code execution flaws in PHP (CVE-2026-109618 and CVE-2026-109619). The patches shipped on October 21. Attackers did not wait for them.

## Why One AI Pentest Run Isn't Enough

DevFeed: [Why One AI Pentest Run Isn't Enough](<https://devfeed.tech/articles/why-one-ai-pentest-run-isn-t-enough-97794.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/why-one-ai-pentest-run-isnt-enough>)

Author: Gunter Ollmann

Published: 2026-10-23T17:46:18Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [reproducibility](<https://devfeed.tech/topics/reproducibility.md>), [trust](<https://devfeed.tech/topics/trust.md>)

Tags: [advanced](<https://devfeed.tech/tags/advanced.md>), [ai](<https://devfeed.tech/tags/ai.md>), [analyst](<https://devfeed.tech/tags/analyst.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-pentest](<https://devfeed.tech/tags/autonomous-pentest.md>), [can](<https://devfeed.tech/tags/can.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [execution](<https://devfeed.tech/tags/execution.md>), [history](<https://devfeed.tech/tags/history.md>), [i](<https://devfeed.tech/tags/i.md>), [industry](<https://devfeed.tech/tags/industry.md>), [input](<https://devfeed.tech/tags/input.md>), [integrity](<https://devfeed.tech/tags/integrity.md>), [isn-t](<https://devfeed.tech/tags/isn-t.md>), [language](<https://devfeed.tech/tags/language.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [large](<https://devfeed.tech/tags/large.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [pentest](<https://devfeed.tech/tags/pentest.md>), [run](<https://devfeed.tech/tags/run.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>)

### AI overview

Because autonomous pentests vary between runs, a single run can miss vulnerabilities or skip methodology steps; repeated runs can broaden findings, but the text says their required number cannot be known because the target's total undiscovered vulnerabilities are unknown. It cites an example where four runs together found 88% of discoverable vulnerabilities although no single run found more than 50%, and notes that varying AI models may increase path diversity. The article recommends reliability mechanisms analogous to human pentesting practices, including focused agents, supervision, human review, and auditable methodology. It also argues that scanners remain essential for efficiently and systematically finding known published flaws: AI can help with iterative probing, but purpose-built scanners can draw on signature libraries and accumulated vulnerability knowledge. Scanners and autonomous pentesters are presented as complementary, not interchangeable.

### Source excerpt

Today's powerful AI and large language model services are non-deterministic. An identical input into the same AI system will yield different results that diverge over execution time. The more vague the input, the longer it needs to reason, and the more advanced the expected output, the greater the variance in responses. In cybersecurity terms, non-determinism translates to missed vulnerability findings, skipped methodology steps, and a fundamental lack of trust in both the quality of findings and the security integrity of the target. If you run one autonomous pentest and trust the results, you're making a bet you can't quantify. The security industry's well-established history of over-claiming AI precision -- something I examined at length in The Cusp of a Virtual Analyst Revolution -- is a useful reminder that "AI says so" has never been a substitute for validated methodology.

## Dissecting the ax-llm CVEs: Why Edge AI Memory and Execution Boundaries Matter

DevFeed: [Dissecting the ax-llm CVEs: Why Edge AI Memory and Execution Boundaries Matter](<https://devfeed.tech/articles/dissecting-the-ax-llm-cves-why-edge-ai-memory-and-execution-boundaries-matter-91065.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/dissecting-the-ax-llm-cves-why-edge-ai-memory-and-execution-boundaries-matter>)

Author: Joe Brinkley

Published: 2026-10-20T14:37:53Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Memory Safety](<https://devfeed.tech/topics/memory-safety.md>), [dma](<https://devfeed.tech/topics/dma.md>), [On-device AI](<https://devfeed.tech/topics/on-device-ai.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [autonomous vehicles](<https://devfeed.tech/topics/autonomous-vehicles.md>), [llama.cpp](<https://devfeed.tech/topics/llama-cpp.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-chip](<https://devfeed.tech/tags/ai-chip.md>), [ai-security](<https://devfeed.tech/tags/ai-security.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [audit](<https://devfeed.tech/tags/audit.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-vehicles](<https://devfeed.tech/tags/autonomous-vehicles.md>), [ax](<https://devfeed.tech/tags/ax.md>), [c](<https://devfeed.tech/tags/c.md>), [c-plus-plus](<https://devfeed.tech/tags/c-plus-plus.md>), [camera](<https://devfeed.tech/tags/camera.md>), [cameras](<https://devfeed.tech/tags/cameras.md>), [chip](<https://devfeed.tech/tags/chip.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cves](<https://devfeed.tech/tags/cves.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [execution](<https://devfeed.tech/tags/execution.md>), [llm](<https://devfeed.tech/tags/llm.md>), [memory](<https://devfeed.tech/tags/memory.md>), [research](<https://devfeed.tech/tags/research.md>)

### AI overview

The excerpt reports seven security defects found in AXERA-TECH's ax-llm C/C++ inference runtime during an audit on AX8850/AX650 edge hardware: five memory-management flaws, audio-path OS command injection, and a null-pointer crash in streaming callbacks. The memory bugs include unchecked writes into NPU memory, integer overflows, an unbounded stack copy, and a double free. The article says maintainers committed fixes across branches, but repository patches do not update deployed firmware; integrators should check their specific branch and BSP against the fixes. The authors withheld exploit PoC code until planned MITRE CVE publication on October 31, 2026. Recommended defenses include validating sizes and tensor bounds, using bounded formatting, avoiding shell invocation for media paths, enforcing C++ ownership and callback checks, and reducing daemon privileges.

### Source excerpt

When the security community talks about AI security, the conversation almost always lands on prompt injection, guardrail evasion, and training data extraction. Those discussions stay at the web application layer. Our research looks lower in the stack. We audit the C and C++ inference runtimes as AI workloads shift directly onto edge silicon. That hardware foundation has critical real-world deployment and industry impact across smart cameras, autonomous vehicles, and industrial systems. Think about a residential security camera at 2:00 AM. It has to decide whether a shadow on a driveway is a person, a pet, or a passing car. It runs that inference locally on a dedicated edge AI chip rather than streaming raw video to the cloud.

## The Token Tax: Why AI Pentesting Gets Expensive Fast

DevFeed: [The Token Tax: Why AI Pentesting Gets Expensive Fast](<https://devfeed.tech/articles/the-token-tax-why-ai-pentesting-gets-expensive-fast-88716.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/the-token-tax-why-ai-pentesting-gets-expensive-fast>)

Author: Gunter Ollmann

Published: 2026-10-16T13:33:18Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI in cybersecurity](<https://devfeed.tech/topics/ai-in-cybersecurity.md>)

Tags: [ai-pentesting](<https://devfeed.tech/tags/ai-pentesting.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [cost](<https://devfeed.tech/tags/cost.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [pentesting](<https://devfeed.tech/tags/pentesting.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>), [token](<https://devfeed.tech/tags/token.md>), [tokens](<https://devfeed.tech/tags/tokens.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

The article argues that AI pentesting can be cheaper for shallow, compliance-focused testing, but token use and testing effort rise sharply with depth. It says current AI systems cannot economically deliver the highest-depth pentesting categories, where human expertise remains important. Buyers should compare solutions against their required testing depth and coverage rather than headline prices.

### Source excerpt

Given unlimited access to advanced AI models, unlimited time, and an unlimited budget to pay for the tokens those models consume, an AI-based pentesting solution should be able to find every discoverable vulnerability in a chosen target, even with today's AI. The problem, obviously, is that nobody has an unlimited budget. And the cost curve for AI-delivered pentesting does not scale the way most buyers assume.

## No Plan Survives First Contact: Scoping the AI-Delivered Pentest

DevFeed: [No Plan Survives First Contact: Scoping the AI-Delivered Pentest](<https://devfeed.tech/articles/no-plan-survives-first-contact-scoping-the-ai-delivered-pentest-65636.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/no-plan-survives-first-contact-scoping-the-ai-delivered-pentest>)

Author: Gunter Ollmann

Published: 2026-10-06T18:59:24Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>)

Tags: [ai-assisted-workflows](<https://devfeed.tech/tags/ai-assisted-workflows.md>), [ai-pentesting](<https://devfeed.tech/tags/ai-pentesting.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [pentesting](<https://devfeed.tech/tags/pentesting.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>)

### AI overview

The article argues that AI-delivered penetration testing requires carefully negotiated scope and ongoing adjustment as findings emerge. It outlines potential AI uses in briefing, document analysis, and scope submission, while arguing that human oversight remains necessary to redirect autonomous testing and keep execution within agreed guardrails.

### Source excerpt

Carefully defining and subsequently adhering to an approved scope is arguably the most important qualitative aspect of any penetration test. A poorly scoped pentest is, at best, wasted time and expense--and at worst, a route to an unrecoverable security incident and permanent data loss. It is therefore a key element of modern pentesting and a critical, underappreciated input for AI-delivered pentesting.

## AI FUD: We've Seen This Movie Before

DevFeed: [AI FUD: We've Seen This Movie Before](<https://devfeed.tech/articles/ai-fud-we-ve-seen-this-movie-before-63385.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/ai-fud-weve-seen-this-movie-before>)

Author: Gunter Ollmann

Published: 2026-10-01T20:02:33Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Cybercrime](<https://devfeed.tech/topics/cybercrime.md>), [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>), [hugging face](<https://devfeed.tech/topics/hugging-face.md>), [openai-hugging-face-incident](<https://devfeed.tech/topics/openai-hugging-face-incident.md>)

Tags: [accelerate](<https://devfeed.tech/tags/accelerate.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-pentesting](<https://devfeed.tech/tags/ai-pentesting.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [exploited](<https://devfeed.tech/tags/exploited.md>), [guardrails](<https://devfeed.tech/tags/guardrails.md>), [hacker](<https://devfeed.tech/tags/hacker.md>), [hacking](<https://devfeed.tech/tags/hacking.md>), [here](<https://devfeed.tech/tags/here.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [internet](<https://devfeed.tech/tags/internet.md>), [intrusion-detection](<https://devfeed.tech/tags/intrusion-detection.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>)

### AI overview

The article argues that current fears about generative AI, including claims of autonomous hacking and other catastrophic harms, echo earlier panic about the Internet. It compares those concerns with 1990s debates and says real risks should be managed through security practices and guardrails developed alongside new capabilities.

### Source excerpt

Every generation believes they're the first to face a technology capable of unraveling civilization. We are not.

## Manual vs. Automated Penetration Testing: Which Should You Choose?

DevFeed: [Manual vs. Automated Penetration Testing: Which Should You Choose?](<https://devfeed.tech/articles/manual-vs-automated-penetration-testing-which-should-you-choose-61917.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/manual-vs.-automated-penetration-testing-which-should-you-choose>)

Author: Claire Bishop

Published: 2026-09-29T18:07:04Z

Content type: comparison

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Application Security](<https://devfeed.tech/topics/application-security.md>), [Mobile Security](<https://devfeed.tech/topics/mobile-security.md>)

Tags: [business-logic-flaws](<https://devfeed.tech/tags/business-logic-flaws.md>), [chaining](<https://devfeed.tech/tags/chaining.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [offensive-security](<https://devfeed.tech/tags/offensive-security.md>), [penetration-testing](<https://devfeed.tech/tags/penetration-testing.md>), [security](<https://devfeed.tech/tags/security.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>)

### AI overview

The article compares manual and automated penetration testing, explaining that they serve different purposes. Manual testing uses expert judgment to identify business logic flaws, connect exposures into attack paths, and validate findings. It is especially suited to complex applications, compliance requirements, and situations where teams need to understand real world risk.

### Source excerpt

Penetration testing is no longer a nice-to-have. With cybersecurity incidents occurring every single day, impacting organizations both large and small, it's now a requirement.

## Automated, Agentic, Autonomous: The AI Pentesting Vocabulary Problem

DevFeed: [Automated, Agentic, Autonomous: The AI Pentesting Vocabulary Problem](<https://devfeed.tech/articles/automated-agentic-autonomous-the-ai-pentesting-vocabulary-problem-59407.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/automated-agentic-autonomous-the-ai-pentesting-vocabulary-problem>)

Author: Gunter Ollmann

Published: 2026-09-24T18:23:49Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Cybersecurity](<https://devfeed.tech/topics/cybersecurity.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Fuzzing/Fuzz testing](<https://devfeed.tech/topics/fuzzing.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Reverse Engineering](<https://devfeed.tech/topics/reverse-engineering.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-pentesting](<https://devfeed.tech/tags/ai-pentesting.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [automated](<https://devfeed.tech/tags/automated.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [penetration-testing](<https://devfeed.tech/tags/penetration-testing.md>)

### AI overview

The article argues that automated, agentic, and autonomous are being used inconsistently in AI penetration-testing marketing. It explains that modern pentesting was already substantially automated and tool-based before current AI products, including the use of reasoning, fuzzing, LLMs, and vendor-trained models.

### Source excerpt

As global penetration testing undergoes its biggest transformation since the invention of PTaaS, offensive security practitioners and buyers are being flooded with marketing jargon and technical comparisons that blur the differences. Prefixes like "Automated," "Agentic," and "Autonomous" are increasingly used interchangeably - and often randomly - to describe features, capabilities, and aspirations of newly AI-powered pentesting products and services. The problem isn't just aesthetic. When vocabulary loses meaning, buyers can't make informed decisions, and vendors have no incentive to be honest about what their products actually do.

## Case Study: How a Bracket Bypassed a Denylist and Exposed a Database Password

DevFeed: [Case Study: How a Bracket Bypassed a Denylist and Exposed a Database Password](<https://devfeed.tech/articles/case-study-how-a-bracket-bypassed-a-denylist-and-exposed-a-database-password-54158.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/case-study-how-a-bracket-bypassed-a-denylist-and-exposed-a-database-password>)

Author: Gaurav Bhosale

Published: 2026-09-17T20:15:47Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [text2sql](<https://devfeed.tech/topics/text2sql.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [SQL](<https://devfeed.tech/topics/sql.md>), [API](<https://devfeed.tech/topics/api.md>), [Reconnaissance](<https://devfeed.tech/topics/recon.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-chatbot](<https://devfeed.tech/tags/ai-chatbot.md>), [ai-integrated](<https://devfeed.tech/tags/ai-integrated.md>), [attack-surface](<https://devfeed.tech/tags/attack-surface.md>), [cobalt-core](<https://devfeed.tech/tags/cobalt-core.md>), [enumeration](<https://devfeed.tech/tags/enumeration.md>), [exploitation](<https://devfeed.tech/tags/exploitation.md>), [security](<https://devfeed.tech/tags/security.md>), [sql](<https://devfeed.tech/tags/sql.md>)

### AI overview

A security case study examines how an AI-powered natural-language-to-SQL assistant allowed an authenticated administrator to retrieve a plaintext database password despite a denylist. It describes reconnaissance, manual testing, API enumeration, and the discovery that the chat interface exposed a raw SQL execution endpoint.

### Source excerpt

Introduction AI features are being bolted onto admin panels faster than security teams can review them. Natural-language-to-SQL assistants are a great example: they're convenient, they demo well, and they quietly reintroduce classic injection-style problems in a new wrapper. This case study walks through a real finding from a recent engagement, where an AI-powered SQL assistant let an authenticated admin retrieve a plaintext database password despite an explicit filter designed to stop exactly that. Client and product details have been generalized throughout.

## From Internet Exposure to Operational Disruption: A Red Team View of Connected Infrastructure

DevFeed: [From Internet Exposure to Operational Disruption: A Red Team View of Connected Infrastructure](<https://devfeed.tech/articles/from-internet-exposure-to-operational-disruption-a-red-team-view-of-connected-infrastructure-54160.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/from-internet-exposure-to-operational-disruption-a-red-team-view-of-connected-infrastructure>)

Author: Saad Nasir

Published: 2026-09-17T00:58:19Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [Risk](<https://devfeed.tech/topics/risk.md>), [Resilience](<https://devfeed.tech/topics/resilience.md>), [Network](<https://devfeed.tech/topics/network.md>), [incident](<https://devfeed.tech/topics/incident.md>), [passwords](<https://devfeed.tech/topics/passwords.md>)

Tags: [cobalt-core](<https://devfeed.tech/tags/cobalt-core.md>), [cyberattack](<https://devfeed.tech/tags/cyberattack.md>), [incident](<https://devfeed.tech/tags/incident.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [network](<https://devfeed.tech/tags/network.md>), [operational](<https://devfeed.tech/tags/operational.md>), [password](<https://devfeed.tech/tags/password.md>), [resilience](<https://devfeed.tech/tags/resilience.md>), [risk](<https://devfeed.tech/tags/risk.md>)

### AI overview

A red-team perspective on how attacks against internet-facing PLCs in water and wastewater systems can cause operational disruption by changing passwords or IP addresses and removing operators' trusted ability to access, manage, verify, and recover systems.

### Source excerpt

When people think about a cyberattack on water infrastructure, they often imagine the most dramatic outcome: a malicious actor changing a setpoint, stopping a pump, manipulating chemical treatment, or interfering with a physical process.

## How Cobalt Brings Real-Time Exploit and Threat Intelligence to Vulnerability Findings

DevFeed: [How Cobalt Brings Real-Time Exploit and Threat Intelligence to Vulnerability Findings](<https://devfeed.tech/articles/how-cobalt-brings-real-time-exploit-and-threat-intelligence-to-vulnerability-findings-54161.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/how-cobalt-brings-real-time-exploit-and-threat-intelligence-to-vulnerability-findings>)

Author: Molly Finn

Published: 2026-09-16T19:21:39Z

Content type: release

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Vulnerabilities](<https://devfeed.tech/topics/vulnerabilities.md>), [common vulnerabilities and exposures](<https://devfeed.tech/topics/common-vulnerabilities-and-exposures.md>), [Threat Research](<https://devfeed.tech/topics/threat-research.md>), [Security research](<https://devfeed.tech/topics/security-research.md>), [High Profile Threats](<https://devfeed.tech/topics/high-profile-threats.md>), [real-time](<https://devfeed.tech/topics/real-time.md>)

Tags: [cve](<https://devfeed.tech/tags/cve.md>), [cybersecurity-services](<https://devfeed.tech/tags/cybersecurity-services.md>), [exploit](<https://devfeed.tech/tags/exploit.md>), [product-innovation](<https://devfeed.tech/tags/product-innovation.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [security-research](<https://devfeed.tech/tags/security-research.md>), [threat-intelligence](<https://devfeed.tech/tags/threat-intelligence.md>), [vulnerability](<https://devfeed.tech/tags/vulnerability.md>)

### AI overview

Cobalt integrated VulnCheck CVE intelligence into its Offensive Security Platform. The integration adds real-time exploit and threat intelligence to pentest findings, including EPSS scores, CISA KEV status, MITRE ATT&CK techniques, and publicly available exploit references. The article also explains that pentests can uncover non-CVE issues such as business logic flaws, access-control weaknesses, and security misconfigurations.

### Source excerpt

Common Vulnerabilities and Exposures (CVE) identify known vulnerabilities in specific components. They don't confirm whether you're exposed or whether attackers are actively exploiting a vulnerability. The additional context from adding reliable threat intelligence and known CVE information to your pentest findings can make the difference between wasting time on research and triage and quickly identifying pressing risks that your business needs to address.

## Pentester Spotlight -- Edu Garcia

DevFeed: [Pentester Spotlight -- Edu Garcia](<https://devfeed.tech/articles/pentester-spotlight-edu-garcia-54162.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/pentester-spotlight-edu-garcia>)

Author: Noelle Hori

Published: 2026-09-15T19:35:53Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Cybersecurity](<https://devfeed.tech/topics/cybersecurity.md>), [Hacking](<https://devfeed.tech/topics/hacking.md>), [Vulnerabilities](<https://devfeed.tech/topics/vulnerabilities.md>), [Software Testing](<https://devfeed.tech/topics/software-testing.md>), [Automation](<https://devfeed.tech/topics/automation.md>), [Authorization](<https://devfeed.tech/topics/authorization.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [apis](<https://devfeed.tech/tags/apis.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [automation](<https://devfeed.tech/tags/automation.md>), [cobalt-core](<https://devfeed.tech/tags/cobalt-core.md>), [core](<https://devfeed.tech/tags/core.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [interview](<https://devfeed.tech/tags/interview.md>), [offensive-security](<https://devfeed.tech/tags/offensive-security.md>), [penetration-testing](<https://devfeed.tech/tags/penetration-testing.md>), [spotlight](<https://devfeed.tech/tags/spotlight.md>), [sql-injection](<https://devfeed.tech/tags/sql-injection.md>), [tools](<https://devfeed.tech/tags/tools.md>), [vulnerabilities](<https://devfeed.tech/tags/vulnerabilities.md>)

### AI overview

An interview with pentester Edu Garcia about his path into cybersecurity, offensive security, blind SQL injection, pentesting methodology, automation, and testing authorization in websites, APIs, and AI systems.

### Source excerpt

The Cobalt Pentester Spotlight highlights the fascinating journey of our Core members. Through an interview style, we share their experiences, background, and insights into the world of an accomplished pentester.

## Beyond Point-in-Time: Continuous Offensive Security with the Cobalt API

DevFeed: [Beyond Point-in-Time: Continuous Offensive Security with the Cobalt API](<https://devfeed.tech/articles/beyond-point-in-time-continuous-offensive-security-with-the-cobalt-api-54157.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/beyond-point-in-time-continuous-offensive-security-with-the-cobalt-api>)

Author: Eugene Revzin

Published: 2026-09-02T20:37:05Z

Content type: release

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [API](<https://devfeed.tech/topics/api.md>), [DevSecOps](<https://devfeed.tech/topics/devsecops.md>), [Vulnerabilities](<https://devfeed.tech/topics/vulnerabilities.md>), [Software Testing](<https://devfeed.tech/topics/software-testing.md>), [GitHub](<https://devfeed.tech/topics/github.md>), [Azure DevOps](<https://devfeed.tech/topics/azure-devops.md>), [jira](<https://devfeed.tech/topics/jira.md>)

Tags: [api](<https://devfeed.tech/tags/api.md>), [azure](<https://devfeed.tech/tags/azure.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [devops](<https://devfeed.tech/tags/devops.md>), [devsecops](<https://devfeed.tech/tags/devsecops.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [github](<https://devfeed.tech/tags/github.md>), [jira](<https://devfeed.tech/tags/jira.md>), [offensive-security](<https://devfeed.tech/tags/offensive-security.md>), [security](<https://devfeed.tech/tags/security.md>), [thought-leadership](<https://devfeed.tech/tags/thought-leadership.md>)

### AI overview

The Cobalt API enables enterprise security programs to automate offensive security workflows, including pentests, findings, DAST scans, and secure code reviews. It connects security data with systems such as Jira, ServiceNow, GitHub, and Azure DevOps to support continuous testing and remediation.

### Source excerpt

For enterprise security programs, the challenge was never running offensive security testing against your applications and networks. It's whether the process and results were manageable at scale and could actually drive faster remediation across every business unit and engineering team in the organization.

## AI Is Changing How We Pentest. Here's What Still Has to Be Human.

DevFeed: [AI Is Changing How We Pentest. Here's What Still Has to Be Human.](<https://devfeed.tech/articles/ai-is-changing-how-we-pentest-here-s-what-still-has-to-be-human-54156.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/ai-is-changing-how-we-pentest-heres-what-still-has-to-be-human>)

Author: Faisal Nur

Published: 2026-08-31T17:49:27Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Security](<https://devfeed.tech/topics/security.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Reconnaissance](<https://devfeed.tech/topics/recon.md>), [Code](<https://devfeed.tech/topics/code.md>), [Parsing](<https://devfeed.tech/topics/parsing.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [autonomous-pentest](<https://devfeed.tech/tags/autonomous-pentest.md>), [offensive-security](<https://devfeed.tech/tags/offensive-security.md>), [parsing](<https://devfeed.tech/tags/parsing.md>), [security](<https://devfeed.tech/tags/security.md>)

### AI overview

The article examines how AI is being incorporated into penetration-testing workflows, including reconnaissance, payload generation, code review, scripting, document summarization, and report drafting. It argues that AI changes the pace of pentesting while human judgment, communication, and understanding of customer needs remain central.

### Source excerpt

Two things are true right now in offensive security.

## 5 Steps: How to Convince Your Boss You Need Cobalt

DevFeed: [5 Steps: How to Convince Your Boss You Need Cobalt](<https://devfeed.tech/articles/5-steps-how-to-convince-your-boss-you-need-cobalt-54155.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/5-steps-how-to-convince-your-boss-you-need-cobalt>)

Author: Luke Doherty

Published: 2026-08-25T22:55:35Z

Content type: opinion

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [Risk](<https://devfeed.tech/topics/risk.md>), [Cloud](<https://devfeed.tech/topics/cloud.md>), [API](<https://devfeed.tech/topics/api.md>)

Tags: [cost-savings](<https://devfeed.tech/tags/cost-savings.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [modernizing-pentesting](<https://devfeed.tech/tags/modernizing-pentesting.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [roi](<https://devfeed.tech/tags/roi.md>), [security](<https://devfeed.tech/tags/security.md>)

### AI overview

A promotional guide explains how to persuade a manager to adopt Cobalt's pentesting as a service. It emphasizes time savings, return on investment, reduced security gaps, continuous testing, and faster remediation compared with point-in-time penetration tests.

### Source excerpt

You've seen Cobalt and pentesting as a service (PTaaS) in action, maybe at your previous job, maybe from a peer who won't stop talking about it, maybe from your own late-night research rabbit hole after another painful pentest cycle. You already know PTaaS beats the old model. You've seen the announcements about Cobalt's new human-directed autonomous pentest offering. The problem is you're the only one in the room who knows about it.

## Cobalt Pentester Spotlight -- Gaurav Bhosale

DevFeed: [Cobalt Pentester Spotlight -- Gaurav Bhosale](<https://devfeed.tech/articles/cobalt-pentester-spotlight-gaurav-bhosale-54159.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/cobalt-pentester-spotlight-gaurav-bhosale>)

Author: Noelle Hori

Published: 2026-08-19T20:42:42Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Cybersecurity](<https://devfeed.tech/topics/cybersecurity.md>), [Hacking](<https://devfeed.tech/topics/hacking.md>), [Vulnerabilities](<https://devfeed.tech/topics/vulnerabilities.md>), [business logic](<https://devfeed.tech/topics/business-logic.md>), [API keys](<https://devfeed.tech/topics/api-keys.md>)

Tags: [business-logic](<https://devfeed.tech/tags/business-logic.md>), [cobalt-core](<https://devfeed.tech/tags/cobalt-core.md>), [cybersecurity](<https://devfeed.tech/tags/cybersecurity.md>), [hacking](<https://devfeed.tech/tags/hacking.md>), [interview](<https://devfeed.tech/tags/interview.md>), [penetration-testing](<https://devfeed.tech/tags/penetration-testing.md>), [vulnerabilities](<https://devfeed.tech/tags/vulnerabilities.md>)

### AI overview

An interview with pentester Gaurav Bhosale about his path into cybersecurity and penetration testing, including his interest in business logic vulnerabilities and sensitive information disclosure.

### Source excerpt

The Cobalt Pentester Spotlight highlights the fascinating journey of our Core members. Through an interview-style format, we share their experiences, background, and insights into the world of an accomplished pentester.

## State of Pentesting in Financial Services and Insurance Industries

DevFeed: [State of Pentesting in Financial Services and Insurance Industries](<https://devfeed.tech/articles/state-of-pentesting-in-financial-services-and-insurance-industries-54163.md>)

Original publisher: [Read original article](<https://www.cobalt.io/blog/state-of-pentesting-in-financial-services-and-insurance-industries>)

Author: Cobalt

Published: 2026-08-18T16:09:35Z

Content type: article

Language: en

Sources: [Cobalt](<https://devfeed.tech/sources/blog-4.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Risk](<https://devfeed.tech/topics/risk.md>), [incident](<https://devfeed.tech/topics/incident.md>), [AI Development](<https://devfeed.tech/topics/ai-development.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [ai-applications](<https://devfeed.tech/tags/ai-applications.md>), [financial-services](<https://devfeed.tech/tags/financial-services.md>), [incident](<https://devfeed.tech/tags/incident.md>), [insurance](<https://devfeed.tech/tags/insurance.md>), [llm-security](<https://devfeed.tech/tags/llm-security.md>), [penetration](<https://devfeed.tech/tags/penetration.md>), [red-teaming](<https://devfeed.tech/tags/red-teaming.md>), [report](<https://devfeed.tech/tags/report.md>), [research](<https://devfeed.tech/tags/research.md>), [risk](<https://devfeed.tech/tags/risk.md>), [security](<https://devfeed.tech/tags/security.md>), [state-of-pentesting](<https://devfeed.tech/tags/state-of-pentesting.md>)

### AI overview

A 2026 snapshot of pentesting in financial services and insurance finds that these sectors test AI applications more than any other sector and report the fewest AI or LLM security incidents. However, remediation performance is average, and the sectors are least likely to expand red-team operations.

### Source excerpt

The financial services and insurance industries are disciplined when it comes to pentesting, with one blind spot. While the financial services and insurance industries test their AI applications more than any other sector and have the fewest AI incidents, they to expand red teaming the least, and their remediation time is only mid-pack. Examined together, these two industries run security like the regulated sectors they are: programmatic, measured, and confident. This state of pentesting snapshot combines five years of Cobalt pentest findings with our 2026 survey of security leaders and practitioners, to profile how these sectors find, fix, and think about security risk. Their teams test AI more than anyone and report the lowest AI incident rate of any industry. Yet underneath that confidence, remediation is only average on the truer half-life measure. And despite testing the most, these two sectors are the least likely to add the one practice built to find what routine testing misses: red teaming. Sources: Cobalt State of Pentesting Report 2026 (~5,000 penetration tests/yr) - 2026 survey of 455 security leaders and practitioners (Emerald Research). Data from financial services and insurance industries are combined due to smaller sample sizes (n=65). The high-risk remediation funnel Every high-risk pentest finding runs this gauntlet, from discovery to closure. The sector resolves a high share of findings, but its half-life sits in the middle of the pack. 9% of findings are high-risk -> 86% of those get resolved -> 46d mean time to remediate -> 55d high-risk half-life 48 % test their AI applications programmatically (highest of any sector) 9 % have had an AI or LLM security incident (lowest of any sector) 17 % plan to increase red team operations (lowest of any sector) Where the financial and insurance industry stands: remediation half-life by sector Chart: the time to remediate half of all high-risk findings. Lower is better. The sector sits mid-pack on half-life (55 d

[Next page](<https://devfeed.tech/sources/blog-4.md?cursor=WyIyMDI2LTA4LTE4VDE2OjA5OjM1KzAwOjAwIiwgImUwYWRlNjMxLTE1NTAtNDY4Zi04MGFmLWQxYjc3MTZmYTZkMyJd>)