threat-intel AI-Generated Patches Fail Half the Time A study by 1Password found that AI-generated patches are significantly flawed, with only around 46% successfully fixing vulnerabilities and many introducing new bugs or requiring code modification. The research, dubbed FLAWED, highlights a concerning asymmetry: AI systems are better at exploiting vulnerabilities than p… Dark Reading · Aug 7, 2026 High UNaipatchingvulnerability
threat-intel Nuclear-Sabotage Malware Benchmark Trips Up Most Frontier AI Models SentinelOne has developed a new benchmark to assess how well frontier AI models can conduct thorough investigations of complex malware, using Fast16 – a malware linked to Iran’s nuclear program – as the test case. The be… SecurityWeek · Jul 23, 2026 Medium IRaimalwareinvestigation
threat-intel OpenAI’s GPT-Red Automates Prompt Injection Testing to Harden GPT-5.6 Sol OpenAI has developed GPT-Red, an automated red-teaming model, to proactively identify and mitigate prompt injection vulnerabilities in its large language models, particularly GPT-5.6 Sol. This model, trained through self… The Hacker News · Jul 16, 2026 High prompt injectionred teamingai security
threat-intel 'Yellow Teams' Are Defining the Future of AI Security A growing trend of ‘yellow teams’ – engineering groups building both attack and defense tools – is emerging as a crucial response to the increasing threat of AI-powered cyberattacks. These teams are using advanced AI mod… Dark Reading · Jul 13, 2026 High aicybersecurityvulnerability
threat-intel Chinese LLMs Broaden the Gap Between Attackers & Defenders This article reports on the emergence of new Chinese AI models, GLM 5.2 and Tulongfeng (Dragon Saber), which are demonstrating strong performance in vulnerability discovery, rivaling leading US models like Opus and GPT-5… Dark Reading · Jul 3, 2026 High CHUSaivulnerabilitychina
threat-intel US lifts export controls on Anthropic’s frontier cybersecurity AI models The U.S. government has lifted export controls on Anthropic’s Fable 5 and Mythos 5 cybersecurity AI models following a ‘jailbreak’ exploit discovered in Fable 5. This marks the first instance of export controls being app… The Record · Jul 1, 2026 Medium USCHaijailbreakexport controls
threat-intel Anthropic Restores Claude Fable 5 After U.S. Lifts Jailbreak-Linked Export Controls Anthropic has restored access to Claude Fable 5 following the U.S. Commerce Department’s lifting of export controls triggered by a jailbreak vulnerability discovered in the model. The controls, implemented in June, restr… The Hacker News · Jul 1, 2026 High USjailbreakaisecurity
threat-intel The AI Token Costs That Can Break Cybersecurity This article highlights a growing concern within the cybersecurity industry: the unexpectedly high costs associated with utilizing AI-powered security platforms, particularly those leveraging generative and agentic AI mo… SecurityWeek · Jun 30, 2026 High aitokenizationcost
threat-intel OpenAI Previews GPT-5.6 Sol With Restricted Access and Stronger Cyber Safeguards OpenAI is releasing a preview of GPT-5.6 Sol, a powerful AI model with enhanced cybersecurity capabilities, to a limited group of companies and the U.S. government. The model is designed for legitimate vulnerability rese… The Hacker News · Jun 27, 2026 High USaicybersecurityvulnerability research
threat-intel OpenAI Refocuses Cybersecurity Efforts on Patching Over Discovery OpenAI is shifting its cybersecurity strategy from solely discovering vulnerabilities to accelerating the process of patching them, recognizing the overwhelming volume of findings generated by AI. The company is deployin… SecurityWeek · Jun 23, 2026 Medium USaicybersecuritypatching
threat-intel Security Community Slams US Ban on Exporting Mythos, Fable The US government recently imposed an export control order restricting access to Anthropic's Claude Fable 5 and Mythos 5 large language models (LLMs) for foreign nationals, citing national security concerns, particularly… Dark Reading · Jun 16, 2026 High USCHllmaiexport control
threat-intel US Cracks Down on Anthropic AI Models Amid Abuse Concerns Anthropic has suspended access to its Fable 5 and Mythos 5 AI models following a US government export control directive, aimed at preventing foreign nationals from utilizing them. This action stems from growing concerns… Dark Reading · Jun 15, 2026 High CHRUUKaicybersecuritythreat intelligence
threat-intel Claude Mythos AI Finds 10,000 High-Severity Flaws in Widely Used Software Anthropic’s Project Glasswing has identified over 10,000 high-severity vulnerabilities in widely used software, primarily through its Claude Mythos Preview AI model. This initiative focuses on proactively identifying and… The Hacker News · May 23, 2026 Critical CVE-2026-5194aivulnerabilitypatching
threat-intel How Dangerous Is Anthropic’s Mythos AI? Anthropic’s Claude Mythos AI model demonstrates a significant capability in identifying software vulnerabilities, prompting concerns about its potential misuse by attackers. While the company initially restricted access… Schneier on Security · May 14, 2026 High UKaivulnerabilitycybersecurity
