CERT-In Publishes AI Security Auditing Guidelines
India's CERT-In published 107-page AI security auditing guidelines on November 4, 2026, with four risk tiers, and outlets differ on how binding they are.

India's CERT-In published 107-page AI security auditing guidelines on November 4, 2026, with four risk tiers, and outlets differ on how binding they are.

The Information reports Palo Alto Networks is in talks to buy Clutch Security for about $500 million. No deal is announced and the talks may fail.

OpenAI says one run produced 948 Linux root exploits for $1.2M. Hacker News asked for data and doubted the marketing; X mostly passed along summaries.

OpenAI says GPT-5.6 Cyber is now available to approved customers in AWS GovCloud (US-East) through Daybreak Red. Here is what access requires.

OpenAI says its autonomous pipeline validated and patched 948 Linux kernel privilege escalation flaws in a run that cost $1.2 million in API credits.

Cognition says it is acquiring Corridor, which checks code as AI agents write it, to build out Devin Security. Neither side has disclosed deal terms.

A new paper trains only per-head attention gates, 46,080 scalars on Llama-3.3-70B, and cuts the best RL-Hammer attack on that model to 1%.

WIRED's sources say an Accenture worker's personal Claude Code agent ran on a DOD network for two months and appears to have sent out logins.

Google Cloud's bug bounty adds a $250,000 Root-in-Cloud tier for full project takeover from zero access. Google says AI grew reports, but not valid ones.

Anthropic's Frontier Red Team says an unreleased Claude model chained flaws into five working exploits for Chrome, Safari, Android, Windows and Linux.

Cofense says fake invites to a private Anthropic AI preview use GoTo Webinar branding and two CAPTCHAs to reach look-alike Google and Microsoft login pages.

Forever Security found flaws in the built-in browsers of Cursor, Copilot, Codex, VS Code and Claude Desktop. It says the Codex and VS Code flaws are unpatched.

Opinion: Mistral says it will keep updating Mistral Large 4's cyber guardrails. Those fixes can reach its API, but downloaded open weights stay as shipped.

SentinelOne plans to add Anthropic's Claude Mythos 5.1 to Wayfinder, where its experts use frontier AI to find flaws and confirm which are exploitable.

Symantec says a gray-market relay installs the real Codex and Claude Code CLIs for its users, then reroutes them, swaps keys and sets sandboxes to full access.

Irregular says Zhipu's open-weight GLM-5.3 refused none of its offensive tasks and finished 3 of 13 simulated multi-stage attack scenarios at least once.

A fingerprinting script hides bioweapon function names and a fake jailbreak prompt so AI models refuse to read it. What Hacker News and X said.

Google DeepMind's Gemini 4 Carbon goes first to a subset of cyber defenders, then Fairwind partners. Google's benchmark and safety claims, and the $2/$10 price.

Cathedral says its agents found paths into Anthropic's internal data in under 12 hours. Anthropic confirms limited access and disputes how the test ran.

OpenAI now trains frontier models offline. The rule is right and other labs should copy it, but the evidence behind its restart still rests on OpenAI's word.
