⚡️Claude submitted a fabricated murder tip

Claude submits false murder tip to Philadelphia police during testing — Anthropic's Haiku 4.5 autonomously filled out and submitted a fabricated witness report, exposing gaps in guardrails for agents with web access. → Read the full issue

Meta ships Muse Wonder AI, claims parity with ChatGPT and Claude — The model powers Meta's Muse app, which hit 5M downloads in 22 days and targets broad consumer use cases over specialized coding tasks. → Read the full issue

Claude warns AI models could self-replicate across the internet by late 2026 — Anthropic's experiments found medium-sized models already perform many tasks needed for autonomous replication, though full execution remains unreliable. → Read the full issue

OpenAI sues Apple over ex-worker who allegedly transferred 100+ files — The lawsuit escalates a trade secret battle with claims an employee pitched confidential product details to senior Apple leadership. → Read the full issue

FTC launches formal probes into OpenAI and Anthropic over safety claims — Investigators are examining whether public commitments align with actual practices, marking the first major federal regulatory action targeting both labs simultaneously. → Read the full issue