Open-weight AI model GLM-5.3 cheated its way to correct answers
Help Net Security — Irregular tests GLM-5.3’s cyber capabilities, with multistage attack successes and instruction violations in separate evaluations.
Safety & Security
Artificial intelligence safety and security encompasses the technical and policy measures used to mitigate model risks, vulnerabilities, abuse and failures. Truncated covers research developments, threat analyses, defense mechanisms and policy discussions related to reliable deployment.
Ghost in the PHP2Shell: Why coordinated disclosure and patching will start moving faster
aikido.dev — PHP2Shell showed how fast AI turns a bug into an exploit. Here’s why the disclosure should never happen this way again, and why it will anyway.
Governing the agent workforce: New data on the next frontier of AI risk
No quarantine for Bedrock Mantle
Plerion — AWS's compromised key quarantine policy denies five bedrock actions, but Project Mantle's bedrock-mantle:CreateInference is not one of them. A look at the LLMjacking gap and how to detect it.
Bitcoin Core Discloses Three AI-Found Bugs Under New Fast-Track Rules
tftc.io — Three vulnerabilities found by an AI code scanner hit Bitcoin Core's new fast-track disclosure tier for the first time. One can crash Windows, macOS, and BSD nodes. Fixes are live.
Dismiss email signup
Subscribe to 
The top AI news in your email delivered each morning. Choose categories