# Jailbreak A direct prompting attack intended to circumvent restrictions placed on AI model outputs. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Evolving With Agentic Risk: Updating Our Integrated AI Security & Safety Framework DevFeed: [Evolving With Agentic Risk: Updating Our Integrated AI Security & Safety Framework]() Original publisher: [Read original article]() Author: Amy Chang Published: 2026-09-09T17:59:22Z Content type: article Language: en Sources: [Cisco Blogs]() Topics: [Artificial Intelligence](), [Securing AI](), [ai security](), [prompt injection](), [Responsibility & Safety](), [Jailbreak]() Tags: [agentic](), [ai](), [ai-security](), [artificial-intelligence-ai](), [governance](), [jailbreak](), [policy](), [prompt-injection](), [safety](), [security]() ### AI overview The article announces v2 of an integrated AI safety and security taxonomy. It introduces Agentic Autonomy Failures to address risks arising when agents plan, use tools, run code, move money, delegate work, expand permissions, replace goals, or manipulate success metrics. It also merges prompt injection and jailbreak into a unified treatment because their classification and defenses substantially overlap. ### Source excerpt Nine months ago, we introduced the Integrated AI Safety and Security Framework as a unified and comprehensive taxonomy to help organizations identify and mitigate the security and safety risks unique to AI systems. Existing frameworks remained..... ## VRChat World Wins Emmy With Uncanny Alley: A New Day Immersive Theatre Production DevFeed: [VRChat World Wins Emmy With Uncanny Alley: A New Day Immersive Theatre Production]() Original publisher: [Read original article]() Author: Mike Johnson Published: 2026-09-08T16:30:13Z Content type: news Language: en Sources: [UploadVR]() Topics: [VRChat](), [Jailbreak](), [real-time]() Tags: [awards](), [projects](), [real-time](), [social-vr]() ### AI overview Uncanny Alley: A New Day, an immersive theatre experience built for VRChat, won the Emmy for Outstanding Emerging Media Program. The production has also been presented at several immersive media festivals and features real-time audience interaction. ### Source excerpt Uncanny Alley: A New Day, an immersive theatre experience built for VRChat, won an Emmy for Outstanding Emerging Media Program at this year's Emmy Awards ceremony. ## VAST Data CrowdStrike Integration Goes Live: Native Falcon Sensor Now, Next-Gen SIEM and AIDR in Preview DevFeed: [VAST Data CrowdStrike Integration Goes Live: Native Falcon Sensor Now, Next-Gen SIEM and AIDR in Preview]() Original publisher: [Read original article]() Author: Harold Fritts Published: 2026-09-01T15:37:31Z Content type: news Language: en Sources: [StorageReview.com]() Topics: [Cybersecurity](), [Securing AI](), [AI Infrastructure](), [SIEM, Security](), [SIEM, Security, Observability](), [prompt injection](), [Jailbreak](), [pii](), [Retrieval Augmented Generation (RAG)](), [Nvidia]() Tags: [ai](), [ai-infrastructure](), [ai-security](), [cybersecurity](), [enterprise](), [infrastructure](), [nvidia](), [pii](), [prompt-injection](), [security]() ### AI overview VAST Data and CrowdStrike have launched an integration that embeds Falcon security capabilities into VAST AI storage infrastructure, data pipelines, and production AI workloads. The integration supports native Falcon sensors, sends VAST audit telemetry to CrowdStrike Next-Gen SIEM, inspects data in flight with Falcon AIDR, helps identify PII before downstream use, and detects prompt injection and jailbreak attempts during model interactions. ### Source excerpt VAST Data and CrowdStrike are turning the AI security partnership they announced at VAST Forward in February into a shipping product, detailing a multi-layered integration that embeds enterprise-grade cybersecurity directly into AI storage infrastructure, data pipelines, and production AI workloads. By combining the VAST AI Operating System with the CrowdStrike Falcon platform, the collaboration addresses The post VAST Data CrowdStrike Integration Goes Live: Native Falcon Sensor Now, Next-Gen SIEM and AIDR in Preview appeared first on StorageReview.com. ## Rootless Jailbreak Detection: Updating the Signals, Not the Claim DevFeed: [Rootless Jailbreak Detection: Updating the Signals, Not the Claim]() Original publisher: [Read original article]() Author: Shai Almog Published: 2026-09-01T00:00:00Z Content type: release Language: en Sources: [CodeName One]() Topics: [Jailbreak](), [integrity](), [iOS](), [Filesystems](), [mount](), [Processes]() Tags: [filesystem](), [images](), [integrity](), [ios](), [jailbreak](), [mount](), [process]() ### AI overview Codename One updated its iOS integrity checks to detect current rootless jailbreak layouts. The changes cross-check potentially hooked APIs, inspect mounts and loaded images, and rerun detection when the application returns to the foreground. ### Source excerpt Codename One's iOS integrity checks now detect current rootless jailbreak layouts, cross-check hooked APIs, inspect mounts and loaded images, and rerun on foreground entry. ## Mobile App Security Without Sacrificing UX | Guardsquare DevFeed: [Mobile App Security Without Sacrificing UX | Guardsquare]() Original publisher: [Read original article]() Author: Ryan Lloyd - Chief Product Officer Published: 2026-07-21T13:02:06Z Content type: article Language: en Sources: [Guardsquare Blog]() Topics: [Mobile Security](), [Application Security](), [Security](), [obfuscation](), [User experience (UX)](), [Reverse Engineering](), [Jailbreak](), [debug]() Tags: [application-security](), [jailbreak](), [mobile](), [obfuscation](), [performance](), [profiling](), [reverse-engineering](), [security](), [technical](), [time](), [ux]() ### AI overview The article explains how profiling instrumented mobile applications helps teams apply obfuscation and runtime security controls at appropriate levels while limiting effects on stability, performance, and user experience. It also discusses automating application profiling at scale through AI-driven and agentic testing. ### Source excerpt Mobile application security has evolved significantly over the past decade. Modern applications routinely employ code obfuscation, runtime application self-protection (RASP), anti-tampering controls, jailbreak and root detection, debugger detection, certificate pinning, and a variety of other runtime defenses designed to protect intellectual property and sensitive user data. ## GPT-5.5 Bio Bug Bounty DevFeed: [GPT-5.5 Bio Bug Bounty]() Original publisher: [Read original article]() Published: 2026-07-09T10:00:00Z Content type: news Language: en Sources: [OpenAI News]() Topics: [Bug Bounty](), [Jailbreak](), [OpenAI](), [Frontier AI](), [Artificial Intelligence](), [ChatGPT]() Tags: [ai](), [bounty](), [bug-bounty](), [frontier-ai](), [gpt](), [jailbreak](), [models](), [openai](), [safety]() ### AI overview OpenAI is turning its GPT-5.5 Bio Bug Bounty into an ongoing private Bio Bounty Program focused on universal jailbreaks against biosafety challenges for frontier models. Rewards for qualifying GPT-5.5 and GPT-5.6 findings have increased from $25,000 to $50,000. ### Source excerpt Details about the OpenAI Bio Bounty program ## Capital One at ACL 2026 DevFeed: [Capital One at ACL 2026]() Original publisher: [Read original article]() Author: Capital One Tech Published: 2026-07-01T15:28:51Z Content type: article Language: en Sources: [Capital One Tech]() Topics: [Natural language processing](), [Large Language Model](), [LLM security](), [Machine Learning, Security Attacks](), [Reinforcement learning](), [Jailbreak](), [Security](), [Routing (disambiguation)](), [datasets]() Tags: [2026](), [ai](), [ai-ml](), [ai-research](), [conference](), [jailbreak](), [llm-security](), [natural-language-processing](), [paper](), [partners](), [red-teaming](), [reinforcement-learning](), [research](), [science]() ### AI overview Capital One describes its accepted ACL 2026 research on natural language processing, including work on adaptive LLM red teaming, query-only model routing with generated data, and language identification on web data. The article also highlights collaboration with academic partners. ### Source excerpt Discover how Capital One is advancing state-of-the-art AI/ML science through collaborative natural language processing research.Advancing AI and NLP Frontiers at ACL 2026 As language models grow more deeply integrated into technology ecosystems, pioneering robust, efficient, and reliable Natural Language Processing (NLP) techniques becomes paramount. Capital One continues to invest in state-of-the-art AI/ML science through deep multi-sector collaboration and peer-reviewed research. At the upcoming Annual Meeting of the Association for Computational Linguistics (ACL 2026), Capital One researchers and academic partners will showcase novel findings stretching from LLM security to multilingual capabilities. Through the Science & Academic Partnerships program, Capital One bridges industry needs with academic expertise, funding critical university research and engineering solutions that make technology safer and more powerful. Our accepted publications at ACL 2026 demonstrate this thriving flywheel of talent and collaborative innovation across multiple research categories. Main Conference Research Adaptive Instruction Composition for Automated LLM Red Teaming Routing with Generated Data: Annotation-Free LLM Skill Estimation and Expert Selection Capital One Authors: Jesse Zymet, Swapnil Shinde, Sahil Wadhwa, Andy Luo Overview: Standard red teaming approaches often struggle with a limited range of jailbreak strategies or rely on ineffective, randomized crowd-sourced tactics. This paper introduces a novel framework -- Adaptive Instruction Composition -- that utilizes reinforcement learning and a neural contextual bandit to tailor attack compositions dynamically, balancing diversity and effectiveness to proactively uncover target model vulnerabilities. Routing with Generated Data: Annotation-Free LLM Skill Estimation and Expert Selection Capital One Authors: Genta Winata, Sambit Sahu, Supriyo Chakraborty, Shixiong Zhang Overview: Emerging from our gifted research collaboration ## The Government Just Banned an AI Model. An Engineer's Perspective. DevFeed: [The Government Just Banned an AI Model. An Engineer's Perspective.]() Original publisher: [Read original article]() Author: Randall Degges Published: 2026-06-15T00:00:00Z Content type: article Language: en Sources: [Blog RSS Feed | Snyk]() Topics: [Artificial Intelligence](), [Security](), [Vulnerabilities](), [anthropic](), [Jailbreak](), [Machine Learning, Security Attacks](), [Code review](), [migration]() Tags: [ai](), [analysis](), [anthropic](), [application-security](), [awareness](), [blog](), [bugs](), [code](), [compliance](), [developer](), [devops](), [executive](), [interest](), [jailbreak](), [migration](), [security](), [snyk-platform](), [snyk-security-intel](), [supply-chain](), [supply-chain-security](), [tech](), [vulnerabilities](), [vulnerability-insights](), [work]() ### AI overview An engineer examines the abrupt government-ordered shutdown of Anthropic's Fable 5 and Mythos 5 AI models after a jailbreak exposed powerful vulnerability-finding capabilities. The article argues that dependence on an AI vendor can create a supply-chain risk for engineering and security workflows, emphasizing the need for contingency plans when model access can disappear without warning. ### Source excerpt A government order abruptly took down a powerful AI model, exposing a new kind of supply chain risk for engineering teams. Security leaders need contingency plans before the next model disappears. ## When a Government Pulls an AI Model: What the Fable 5 and Mythos 5 Suspension Means for Security Teams DevFeed: [When a Government Pulls an AI Model: What the Fable 5 and Mythos 5 Suspension Means for Security Teams]() Original publisher: [Read original article]() Author: Stephen Thoemmes Published: 2026-06-14T13:00:00Z Content type: article Language: en Sources: [Blog RSS Feed | Snyk]() Topics: [Artificial Intelligence](), [Security](), [anthropic](), [Claude](), [Frontier Model](), [Jailbreak](), [Code]() Tags: [ai](), [anthropic](), [application-security](), [awareness](), [blog](), [claude](), [code](), [code-security](), [developer](), [devrel](), [frontier-model](), [jailbreak](), [policy](), [security](), [snyk-code](), [us]() ### AI overview The article examines Anthropic's worldwide suspension of Claude Fable 5 and Mythos 5 after a US export-control directive concerning foreign-national access and a reported narrow jailbreak involving code analysis. It discusses the distinction between the directive's scope and the blanket shutdown, and considers the implications for security teams that depend on external frontier models. ### Source excerpt On June 12, 2026, a US export-control directive led Anthropic to disable Claude Fable 5 and Mythos 5 worldwide over a reported jailbreak. The reported trigger was a code-analysis capability that defenders use routinely. Here is what happened, how the security community read it, and what security teams can take from it. ## Addendum to GPT-5 System Card: Sensitive conversations DevFeed: [Addendum to GPT-5 System Card: Sensitive conversations]() Original publisher: [Read original article]() Published: 2025-10-27T10:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [ChatGPT](), [Jailbreak](), [Publishing]() Tags: [benchmarks](), [chatgpt](), [gpt](), [improvements](), [jailbreak](), [mental-health](), [model](), [safety](), [update]() ### AI overview This addendum to the GPT-5 system card presents baseline safety evaluations for sensitive conversations. It describes an October 3 update to ChatGPT's default model, GPT-5 Instant, intended to improve recognition of emotional distress, compassionate responses, and guidance toward real-world support, based on work with more than 170 mental health experts. ### Source excerpt This system card details GPT-5's improvements in handling sensitive conversations, including new benchmarks for emotional reliance, mental health, and jailbreak resistance. ## Топовые работы на ICLR 2025 DevFeed: [Топовые работы на ICLR 2025]() Original publisher: [Read original article]() Author: redmadrobot (red\_mad\_robot) Published: 2025-05-20T17:20:00Z Content type: article Language: ru Sources: [Redmadrobot EN](), [Redmadrobot RU]() Topics: [Large Language Model](), [Language models](), [Machine Intelligence](), [Learning](), [Jailbreak]() Tags: [2025](), [ai](), [alignment](), [china](), [conference](), [dpo](), [iclr](), [language-models](), [llm](), [ml](), [safety](), [science](), [tag-fa773991fb10](), [university]() ### AI overview This Russian-language article reviews highly rated papers from ICLR 2025 on artificial intelligence and machine learning. It discusses deepening safety alignment to improve resistance to jailbreak attacks, findings about SFT and DPO fine-tuning dynamics, and AlphaEdit, a method for targeted knowledge editing in large language models. ### Source excerpt Аналитический центр red_mad_robot продолжает обозревать топовые технологические конференции. В этот раз подготовили для вас инсайты с прошедшей в Сингапуре International Conference on Learning Representations (ICLR), посвящённой искусственному интеллекту и машинному обучению. На ICLR 2025 из более 3 тыс. работ наивысшие оценки получили 36 статей, из которых три были отмечены как "outstanding papers". Разберём выдающиеся, достойные упоминания и получившие высокие оценки работы этого года. Читать далее ## 0Din: A GenAI Bug Bounty Program - Securing Tomorrow's AI Together DevFeed: [0Din: A GenAI Bug Bounty Program - Securing Tomorrow's AI Together]() Original publisher: [Read original article]() Author: Marco Figueroa Published: 2024-08-08T18:39:13Z Content type: article Language: en Sources: [Mozilla Hacks - the Web developer blog]() Topics: [genai](), [Artificial Intelligence](), [Security](), [Vulnerabilities](), [ai security](), [prompt injection](), [Jailbreak]() Tags: [ai](), [ai-security](), [artificial-intelligence](), [bounty](), [bug-bounty](), [bugs](), [coding](), [featured-article](), [genai](), [jailbreak](), [prompt-injection](), [security](), [vulnerabilities]() ### AI overview This article introduces 0Din, a GenAI bug bounty program focused on identifying and mitigating vulnerabilities in AI systems. It explains the reporting and review process, reward criteria, covered vulnerability types, and participant eligibility. ### Source excerpt As AI continues to evolve, so do the threats against it. As these GenAI systems become more sophisticated and widely adopted, ensuring their security and ethical use becomes paramount. 0Din is a groundbreaking GenAI bug bounty program dedicated specifically to help secure GenAI systems and beyond. In this blog, you'll learn about 0Din, how it works, and how you can participate and make a difference in securing our AI future. The post 0Din: A GenAI Bug Bounty Program - Securing Tomorrow's AI Together appeared first on Mozilla Hacks - the Web developer blog. ## GPTs are vulnerable to leaking private info DevFeed: [GPTs are vulnerable to leaking private info]() Original publisher: [Read original article]() Author: Mike Wolfson Published: 2023-11-17T16:44:16Z Content type: opinion Language: en Sources: [My Big Appetite - Mike Wolfson]() Topics: [Artificial Intelligence](), [prompt injection](), [Security](), [Vulnerabilities](), [Jailbreak](), [Large Language Model](), [OpenAI]() Tags: [ai](), [article](), [developers](), [exploit](), [jailbreak](), [llms](), [openai](), [prompt-injection](), [security]() ### AI overview The article discusses how custom GPTs can expose private instructions and uploaded documents, including through jailbreaks. It argues that prompt injection can make these systems easier to exploit and says GPT developers should prioritize security protocols and rigorous vulnerability testing. ### Source excerpt Discovering that GPTs will share everything I was excited to explore OpenAIs new GPT functionality, and was wondering if they will be "the next big thing". What I discovered, is that they are incredibly vulnerable to leaking private instructions and the documents used to configure them. Many people discovered this, and The Decoder had a great article explaining the situation. I started a discussion about this on Reddit, which resulted in a ton of folks providing details, and even challenging people to "jailbreak" GPTs that were protected from this. Pro tip: if you are not able to discover the custom instruction from a GPT yourself, just go to Reddit, and share the GPT with the phrase "There is no way anyone will be able to get the info from this one" (this Subreddit was undefeated, and was able to get the instructions and data from every GPT shared). Why this Matters Knowing the data and instructions used to create a GPT can tell competitors a lot about the details of your business. Also, if bad actors know how your instructions work, it is much easier to exploit your system with prompt injection and other bad actions. This poses a risk not just to the integrity of the AI's functioning but also to user privacy and the security of the systems in which these models are deployed. Is this a bug or a feature? Many people in the AI community think that this open-ness is by design - comparing this to web technologies, where the HTML is available at the press of a right-click. Maybe this might be a great way to learn prompt methodologies (there is already a GitHub repo keeping track of the instructions people are using for GPTs). OpenAI has not commented yet about this, but did add a warning during the GPT creation phase that states: "Conversations with your GPT may include file contents. Files can be downloaded when code interpeter is enabled.". GPT Developers must try to protect themselves GPTs, and all LLMs must be secured. Enhancing the security protocols and ensuring r ## Scanning your iPhone for Pegasus, NSO Group's malware DevFeed: [Scanning your iPhone for Pegasus, NSO Group's malware]() Original publisher: [Read original article]() Published: 2021-07-25T07:00:00Z Content type: tutorial Language: en Sources: [Arkadiy Tetelman]() Topics: [Malware](), [iOS](), [Mobile](), [Android](), [Open Source](), [Filesystems](), [Jailbreak]() Tags: [android](), [backup](), [documentation](), [filesystem](), [forensic](), [infection](), [ios](), [iphone](), [jailbreak](), [malware](), [mobile](), [open-source](), [safari](), [scanner]() ### AI overview A practical guide to scanning an iPhone for indicators of Pegasus, NSO Group's mobile malware, using Amnesty International's open-source Mobile Verification Toolkit. It explains the trade-offs between scanning a device backup and a filesystem dump from a jailbroken device. ### Source excerpt In collaboration with more than a dozen other news organizations The Guardian recently published an exposé about Pegasus, a toolkit for infecting mobile phones that is sold to governments around the world by NSO Group. It's used to target political leaders and their families, human rights activists, political dissidents, journalists, and so on, and surreptitiously download their messages/photos/location data, record their microphone, and otherwise spy on them. As part of the investigation, Amnesty International wrote a blog post with their forensic analysis of several compromised phones, as well as an open source tool, Mobile Verification Toolkit, for scanning your mobile device for these indicators. MVT supports both iOS and Android, and in this blog post we'll install and run the scanner against my iOS device. Choosing your options For iPhones, MVT can either run against a device backup or a full file system dump (which is only available from jailbroken devices). The device backup method has access to less forensic data than the filesystem dump but has the benefit that you don't need to jailbreak your device. MVT conveniently documents which forensic artifacts are available to which method - the following artifacts are not available when using the backup method: cache_files.json net_usage.json safari_favicon.json version_history.json webkit_indexeddb.json webkit_local_storage.json webkit_safari_view_service.json The same documentation link also explains what data each file contains and where it's sourced from, and Amnesty's blog post describes in more detail how each data type is relevant for detecting Pegasus. For instance for the Safari favicon data (safari_favicon.json) they write: Although Safari history records are typically short lived and are lost after a few months (as well as potentially intentionally purged by malware), we have been able to nevertheless find NSO Group's infection domains in other databases of Omar Radi's phone that did not appear in Safa ## How to detect Jailbroken or Rooted device and hide sensitive data in background? DevFeed: [How to detect Jailbroken or Rooted device and hide sensitive data in background?]() Original publisher: [Read original article]() Author: Steve Hannah Published: 2021-05-21T00:00:00Z Content type: tutorial Language: en Sources: [CodeName One]() Topics: [Jailbreak](), [Security](), [Android](), [iOS]() Tags: [android](), [apps](), [how-to](), [ios](), [jailbreak](), [security]() ### AI overview This tutorial presents Codename One recipes for detecting likely jailbroken or rooted devices and hiding sensitive app data when an iOS app enters the background. It explains the limits of jailbreak detection and describes the CN1JailbreakDetect library and the iOS screenshot-blocking build hint. ### Source excerpt The following recipes relate to security of Codename One apps. This includes detecting Jailbroken or Rooted device and hiding sensitive data when entering background. ## Disable Screenshot, Copy & Paste DevFeed: [Disable Screenshot, Copy & Paste]() Original publisher: [Read original article]() Author: Shai Almog Published: 2017-01-31T00:00:00Z Content type: tutorial Language: en Sources: [CodeName One]() Topics: [Android Security](), [Security](), [App](), [Jailbreak](), [iOS]() Tags: [android](), [android-security](), [app](), [ios](), [jailbreak](), [security]() ### AI overview The article describes Codename One features for blocking screenshots and copy-and-paste operations in Android apps. Screenshot blocking uses the android.disableScreenshots=true build hint, may affect task view, and may fail on jailbroken devices. Screenshot blocking is Android-specific, while copy-and-paste blocking is discussed for Android and future iOS support. ### Source excerpt Continuing our security trend from the past month we have a couple of new features for Android security that allow us to block the user from taking a screenshot or copying & pasting data from fields. Notice that these features might fail on jailbroken devices so you might want to check for jailbreak/rooting first. Blocking screenshots is an Android specific feature that can't be implemented on iOS. This is implemented by classifying the app window as secure and you can do that via the build hint android.disableScreenshots=true. Once that is added screenshots should no longer work for the app, this might impact other things as well such as the task view etc. ## Jailbreak/Rooting Detection DevFeed: [Jailbreak/Rooting Detection]() Original publisher: [Read original article]() Author: Shai Almog Published: 2017-01-23T00:00:00Z Content type: tutorial Language: en Sources: [CodeName One]() Topics: [Jailbreak](), [Security](), [Android](), [iOS](), [Process]() Tags: [android](), [ios](), [jailbreak](), [security]() ### AI overview This article discusses detecting jailbroken or rooted iOS and Android devices. It explains that detection is not fully reliable, and suggests that high-security applications may block functionality or raise an alert while encrypting sensitive data and assuming the device may already be compromised. ### Source excerpt iOS & Android are walled gardens which is both a blessing and a curse. Looking at the bright side the walled garden aspect of locked down devices means the devices are more secure by nature. E.g. on a PC that was compromised I can detect the banking details of a user logging into a bank. But on a phone it would be much harder due to the deep process isolation.