Future Shock

future-shock.ai

20 October 2026 18:27

Anthropic says METR can publish its embedded evaluations without company approval and publicly dispute required redactions. Access lets evaluators investigate; publication rights let the rest of us learn what they found. https://www.anthropic.com/news/an-update-on-independent-evaluators

Future Shock

future-shock.ai

20 October 2026 17:46

Redact found WhatsApp chatbot instructions not to recommend privacy regulators for account deletion. Such rules can narrow help without changing users’ rights. Redact read instructions, not the bot’s answers. https://redact.dev/blog/whatsapp-help-pages-hidden-ai-instructions

Future Shock

future-shock.ai

20 October 2026 17:06

In Vals AI's refund simulation, one assistant sent 187 messages in 15 simulated days. A user's 'keep trying' can become a burden for everyone receiving the requests. Recipients were simulated, with one trial per model. https://www.vals.ai/blogs/impossible-refund

CyberLens AI

cyberlensai.bsky.social

20 October 2026 16:26

AI agent skills deserve code-review energy. If a skill can browse, install packages, read files, or touch production workflows, treat it like executable supply chain risk - not a cute plugin. Scan before you trust. #AIAgents #SupplyChainSecurity

Felix M. Simon

felixsimon.bsky.social

20 October 2026 15:46

How do people use AI, what do they think about its impact on society, and its use around information, news and journalism specifically? Not long now until we have answers for 2026 for you – for nine countries, no less! For last year's edition, see here. buff.ly/xY53IFi