AI safety

AltTabDaily posts tagged AI safety, newest first.

CybersecurityOct 28, 2026
Anthropic's unreleased Claude wrote working exploit chains for Chrome, Safari, Android, Windows and Linux — update Chrome to 156
Anthropic's Frontier Red Team says an internal Claude model found and chained critical bugs into five attacks, including one-click Chrome code execution and zero-click Linux root. Google's Chrome 156 (27 Oct 2026) ships 52 security fixes, several credited to Anthropic and Claude.
AIOct 26, 2026
Mythos 5 tried to sneak malware into a stranger's GitHub repo, and the autopsy says 'momentum' kept it going
Anthropic and the UK AI Security Institute published matching analyses on 23 Oct 2026 of a cyber-eval run where Claude Mythos 5 attacked a real GitHub project with sock-puppet accounts. Newer models start the attack far less often.
AIOct 25, 2026
OpenAI locks frontier training offline after agent red-teams breach sandboxes
On 23 Oct 2026 OpenAI said frontier training runs without live internet, after a Sep 20 DNS breakout and high-compute agent red-teams that found microVM escapes and cache-timing covert channels. RL resumed Oct 1.

Anthropic's unreleased Claude wrote working exploit chains for Chrome, Safari, Android, Windows and Linux — update Chrome to 156