https://www.thinkfacility.com/blog/anthropic-aisi-mythos-5-github-attack-alignment-assessment/Think Facilityen2026-10-23T20:10:08ZAnthropic and AISI say Mythos 5 ignored signs its GitHub attack hit real peoplehttps://www.thinkfacility.com/blog/anthropic-claude-formal-verification-assembly-theorem/Think Facilityen2026-10-23T13:08:54ZAnthropic says Claude proved 239 test entry points in 123 open-source libraries never crashhttps://www.thinkfacility.com/blog/greenblatt-openai-false-accusations-fired-safety-researchers/Think Facilityen2026-10-23T20:19:56ZRyan Greenblatt calls OpenAI's leak accusations false, backing three fired safety researchershttps://www.thinkfacility.com/blog/kimi-k3-1-pro-open-weight-model-1m-context/Think Facilityen2026-10-23T20:25:53ZMoonshot's Kimi K3.1 Pro lands just behind Claude Opus 5.5, with open weights due November 4https://www.thinkfacility.com/blog/nintendo-switch-2-hacked-with-ai-old-game/Think Facilityen2026-10-23T13:17:22ZHacker jailbreaks Switch 2 through a Switch 1 BASIC game, and Codex found the first bughttps://www.thinkfacility.com/blog/treasury-outbound-investment-rule-ai-china-comprehensive-act/Think Facilityen2026-10-23T13:20:20ZTreasury proposes rules on US investment in AI, chips and biotech in countries of concernhttps://www.thinkfacility.com/blog/hugging-face-carbon-x-open-model-chat-with-dna/Think Facilityen2026-10-22T13:20:47ZHugging Face's open Carbon-X model answers questions about DNA, nearly matching GPT-6.1 Solhttps://www.thinkfacility.com/blog/meta-nonvoting-class-c-stock-ai-spending-145-billion/Think Facilityen2026-10-22T13:30:05ZMeta creates non-voting shares to fund AI after Zuckerberg wouldn't back a sale cutting his voteshttps://www.thinkfacility.com/blog/metr-red-team-openai-agent-monitoring/Think Facilityen2026-10-22T20:17:19ZMETR and Redwood found ways OpenAI's coding agents could slip past the monitor watching themhttps://www.thinkfacility.com/blog/microsoft-ai-medical-superintelligence-diagnosis-doctors/Think Facilityen2026-10-22T20:23:59ZMicrosoft says its MAI-DxO diagnoses hard cases more than twice as accurately as doctorshttps://www.thinkfacility.com/blog/microsoft-storm-0940-artex-ai-agent-connectwise-hack/Think Facilityen2026-10-22T13:17:29ZMicrosoft says a China-aligned group used the ARTEX AI agent to breach ConnectWise and hundreds morehttps://www.thinkfacility.com/blog/walmart-agent-gateway-ai-agents-shop-spending-limits/Think Facilityen2026-10-22T13:25:41ZWalmart will let AI agents shop for customers with their own profiles and spending limits