# Adversarial attacks Adversarial attacks are attacks that exploit vulnerabilities in AI and machine-learning systems, including by manipulating inputs or training data. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## AI Model Risk Intelligence Know Which Models You Can Trust Before You Deploy DevFeed: [AI Model Risk Intelligence Know Which Models You Can Trust Before You Deploy]() Original publisher: [Read original article]() Author: Ranko Cupovic Published: 2026-08-04T04:00:00Z Content type: article Language: en Sources: [Blog RSS Feed | Snyk]() Topics: [Artificial Intelligence](), [Adversarial attacks](), [Security](), [asr](), [Benchmark](), [Deployment](), [pii](), [Code]() Tags: [ai](), [application-security](), [asr](), [attacks](), [awareness](), [benchmark](), [blog](), [deployment](), [developer](), [devops](), [enablement](), [model](), [pii](), [safety](), [security](), [snyk-platform](), [snyk-security-intel]() ### AI overview Evo's AI model risk score combines attack success rate and attack impact into a 0-1000 score. It is based on adversarial testing against standard system-prompt hardening and breaks risk down by attacker goals such as PII extraction, system-prompt extraction, and insecure code generation to support deployment decisions. ### Source excerpt AI model risk depends on how a model is deployed. Learn how Evo combines adversarial testing, attack impact, and deployment context to help teams compare models and enforce policy. ## Google Workspace's continuous approach to mitigating indirect prompt injections DevFeed: [Google Workspace's continuous approach to mitigating indirect prompt injections]() Original publisher: [Read original article]() Author: Kimberly Samra (noreply@blogger.com) Published: 2026-04-02T16:00:00Z Content type: opinion Language: en Sources: [Google Online Security]() Topics: [prompt injection](), [Google](), [Security](), [Adversarial attacks](), [Machine Learning, Security Attacks](), [Machine learning](), [Artificial Intelligence](), [Google AI]() Tags: [ai](), [attacks](), [automated](), [discovery](), [google](), [llms](), [machine-learning](), [none](), [prompt](), [prompt-injection](), [red-teaming](), [research](), [safety](), [security](), [testing]() ### AI overview Google describes its ongoing approach to mitigating indirect prompt injection attacks against Workspace with Gemini. The approach includes discovering new attack vectors, human and automated red-teaming, and collaboration with external researchers through the Google AI Vulnerability Rewards Program. ### Source excerpt Posted by Adam Gavish, Google GenAI Security Team Indirect prompt injection (IPI) is an evolving threat vector targeting users of complex AI applications with multiple data sources, such as Workspace with Gemini. This technique enables the attacker to influence the behavior of an LLM by injecting malicious instructions into the data or tools used by the LLM as it completes the user's query. This may even be possible without any input directly from the user. IPI is not the kind of technical problem you "solve" and move on. Sophisticated LLMs with increasing use of agentic automation combined with a wide range of content create an ultra-dynamic and evolving playground for adversarial attacks. That's why Google takes a sophisticated and comprehensive approach to these attacks. We're continuously improving LLM resistance to IPI attacks and launching AI application capabilities with ever-improving defenses. Staying ahead of the latest indirect prompt injection attacks is critical to our mission of securing Workspace with Gemini. In our previous blog "Mitigating prompt injection attacks with a layered defense strategy", we reviewed the layered architecture of our IPI defenses. In this blog, we'll share more detail on the continuous approach we take to improve these defenses and to solve for new attacks. New attack discovery By proactively discovering and cataloging new attack vectors through internal and external programs, we can identify vulnerabilities and deploy robust defenses ahead of adversarial activity. Human Red-Teaming Human Red-Teaming uses adversarial simulations to uncover security and safety vulnerabilities. Specialized teams execute attacks based on realistic user profiles to exploit weaknesses, coordinating with product teams to resolve identified issues. Automated Red-Teaming Automated Red-Teaming is done via dynamic, machine-learning-driven frameworks to stress-test environments. By algorithmically generating and iterating on attack payloads, we can mimi ## AprielGuard: A Guardrail for Safety and Adversarial Robustness in Modern LLM Systems DevFeed: [AprielGuard: A Guardrail for Safety and Adversarial Robustness in Modern LLM Systems]() Original publisher: [Read original article]() Author: Jaykumar Kasundra Published: 2025-12-23T14:07:35Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [Large Language Model](), [prompt injection](), [Adversarial attacks](), [Security](), [Chain-of-thought](), [Architecture & Design](), [Training AI Models]() Tags: [agent](), [agents](), [attacks](), [chain-of-thought](), [latency](), [llm](), [model](), [prompt-injection](), [reasoning](), [safety-security](), [security](), [training]() ### AI overview AprielGuard is an 8B-parameter safety and security safeguard model for modern LLM systems. It detects 16 categories of safety risks and a broad range of adversarial attacks, including prompt injection, jailbreaks, chain-of-thought corruption, context hijacking, memory poisoning, and multi-agent exploit sequences. It supports standalone prompts, multi-turn conversations, and agentic workflows containing tool calls, reasoning traces, memory, and system context. The model offers reasoning and non-reasoning modes for explainable or low-latency classification. ### Source excerpt In this work, we introduce AprielGuard, an 8B parameter safety-security safeguard model designed to detect: - 16 categories of safety risks, spanning toxicity, hate, sexual content, misinformation, self-harm, illegal activities, and more. - Wide range of adversarial attacks, including prompt injection, jailbreaks, chain-of-thought corruption, context hijacking, memory poisoning, and multi-agent exploit sequences. ## Continuously hardening ChatGPT Atlas against prompt injection DevFeed: [Continuously hardening ChatGPT Atlas against prompt injection]() Original publisher: [Read original article]() Published: 2025-12-22T00:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [prompt injection](), [Security](), [ai security](), [ChatGPT](), [OpenAI](), [browser](), [Adversarial attacks](), [Reinforcement learning]() Tags: [agent](), [agentic](), [ai](), [ai-security](), [atlas](), [attacks](), [chatgpt](), [hardening](), [prompt-injection](), [reinforcement-learning](), [security](), [techniques](), [tool](), [web]() ### AI overview OpenAI describes how it is continuously hardening ChatGPT Atlas against prompt injection. Automated red teaming powered by reinforcement learning helps discover novel attacks and exploits so the browser agent's defenses can be improved before attacks spread in the wild. ### Source excerpt OpenAI is strengthening ChatGPT Atlas against prompt injection attacks using automated red teaming trained with reinforcement learning. This proactive discover-and-patch loop helps identify novel exploits early and harden the browser agent's defenses as AI becomes more agentic. ## Why Threat Modeling Is Now Even More Critical for AI-Native Applications DevFeed: [Why Threat Modeling Is Now Even More Critical for AI-Native Applications]() Original publisher: [Read original article]() Author: Snyk Team Published: 2025-11-20T05:00:00Z Content type: article Language: en Sources: [Blog RSS Feed | Snyk]() Topics: [Security](), [ai security](), [Securing AI](), [Artificial Intelligence](), [Large Language Model](), [Continuous Deployment (CD)](), [Adversarial attacks]() Tags: [ai](), [ai-security](), [architecture](), [architectures](), [attacks](), [blog](), [continuous-deployment](), [engineering](), [executive](), [interest](), [language-models](), [large-language-models](), [security]() ### AI overview AI-native applications require continuous, adaptive threat modeling because their large language models, autonomous agents, evolving data flows, and runtime-integrated tools create dynamic and unpredictable security risks. The article explains why traditional manual, static, point-in-time assessments become stale and fail to scale, and highlights new attack surfaces such as data poisoning, model inversion, and adversarial attacks. ### Source excerpt AI-native applications demand a security approach as dynamic as they are. Traditional threat modeling is no longer enough. Discover the shift to continuous, adaptive threat modeling for AI security. ## The New Threat Landscape: AI-Native Apps and Agentic Workflows DevFeed: [The New Threat Landscape: AI-Native Apps and Agentic Workflows]() Original publisher: [Read original article]() Author: Snyk Team Published: 2025-06-17T23:00:00Z Content type: article Language: en Sources: [Blog RSS Feed | Snyk]() Topics: [Application Security](), [Artificial Intelligence](), [prompt injection](), [Security](), [Adversarial attacks](), [Large Language Model](), [Model Context Protocol](), [Authentication](), [Authorization]() Tags: [agentic-ai](), [ai](), [ai-agents](), [awareness](), [blog](), [devops](), [enterprise](), [executive](), [external](), [open-source-vulnerabilities](), [prompt-injection](), [sdlc](), [security](), [snyk-platform]() ### AI overview The article examines the expanding threat landscape around AI-native applications and agentic workflows. It discusses risks including data poisoning, prompt injection, open-source vulnerabilities, and identity gaps, and argues that traditional application security practices need stronger visibility and policy enforcement for AI systems. ### Source excerpt As AI-native apps and agentic workflows expand the attack surface, new threats like prompt injection and data poisoning emerge. Learn why traditional AppSec falls short and how to secure your AI systems in this new threat landscape. ## Application and AI roundup - October DevFeed: [Application and AI roundup - October]() Original publisher: [Read original article]() Author: Adam Published: 2023-11-09T00:00:00Z Content type: article Language: en Sources: [Shostack & Friends Blog]() Topics: [Application Security](), [Artificial Intelligence](), [Security](), [Vulnerabilities](), [Adversarial attacks](), [Large Language Model](), [browser](), [okta](), [solarwinds]() Tags: [ai](), [appsec](), [article](), [attacks](), [browser](), [large-language-models](), [okta](), [security](), [solarwinds](), [vulnerabilities]() ### AI overview An October roundup covering the SEC's case against SolarWinds and its CISO, research on vulnerabilities in large language models and adversarial attacks, AI policy and model behavior, and threat-modeling issues involving browser privacy, Okta's support system, and bug hunting. ### Source excerpt Exciting news from the SEC, lots of AI, and lots of threat modeling. ## Threat Modeling Thursday: Machine Learning DevFeed: [Threat Modeling Thursday: Machine Learning]() Original publisher: [Read original article]() Author: Adam Published: 2020-01-02T00:00:00Z Content type: opinion Language: en Sources: [Shostack & Friends Blog]() Topics: [Machine learning](), [Security](), [Adversarial attacks](), [Reverse Dependencies]() Tags: [analysis](), [dependencies](), [machine-learning](), [ml-supply-chain](), [security](), [sensitive-data]() ### AI overview The article reviews Microsoft documents on threat modeling AI and machine learning systems, including their engineering challenges, categorized attacks, failure modes, and dependencies. It supports combining attacks and failures into a framework while arguing that additional ways to distinguish different security concerns are needed. ### Source excerpt For my first blog post of 2020, I want to look at threat modeling machine learning systems.