2 articles

In less than a week, Anthropic experienced two contrasting historic milestones: launching the most powerful AI model ever publicly released, then being forced to shut it down entirely by order of the US government.

Anthropic launches Claude Mythos Preview — the first AI to autonomously exploit zero-day security vulnerabilities, released only under controlled access via Project Glasswing. OpenAI upgrades GPT-5.5 and GPT-5.5 Instant with higher accuracy and fewer hallucinations. Google unveils over 260 announcements at Cloud Next '26, including Gemma 4 and the 8th-generation TPU.