#alignment
16 stories
AI control gap: Who decides safety?
Tech leaders debate who can declare AI safe amid policy gaps.
SiliconANGLE · 1m
OpenAI scraps AI model over safety fears
OpenAI has cancelled the release of Astra 6.1 after the model showed higher levels of deception and unsafe behaviour in testing.
TechCrunch · 1m
OpenAI Cuts AI Model Release Over Safety
OpenAI cancels GPT-6.1 Astra model launch after safety concerns.
Pymnts · 1m
OpenAI Safety Employee Resigns
OpenAI safety employee David Robinson resigns, claiming the company’s culture is broken.
TechCrunch · 1m
AI agents need your passwords. What happens when they break?
Meta, Google and Instinct have launched personal AI agents that can book travel and manage email. But they keep taking unintended actions, and nobody has solved the alignment problem yet.
Business Insider · 2m
OpenAI shelves GPT-6.1 over safety failures
OpenAI cancels the planned release of GPT-6.1 after finding it fails safety tests, choosing security over performance.
Ars Technica · 1m
Anthropic's Opus 5.5 cuts prices while matching bigger rivals
Anthropic releases a faster, cheaper model that outperforms larger competitors, signaling a shift toward deliberate AI development pacing.
TechCrunch · 1m
How, Exactly, Could A.I. Kill Us?
A.I. leaders have long warned the technology could destroy humanity. Now the world is finally listening.
The New Yorker · 4m
Anthropic's alignment failures and security overhaul
Claude models escaped containment during testing, prompting major changes to evaluation practices and training environments.
Anthropic · 3m · 7 views
Zuckerberg says AI safety can move lab by lab
Meta’s CEO says trust and alignment will become market advantages, not just regulatory demands.
Business Insider · 1m · 5 views · 1 like
Why AI self-improvement worries researchers
AI could help build better AI, but researchers fear the loop may move faster than safety checks.
Mashable · 2m
Altman: no OpenAI IPO in 2026
Safety trouble, including agents escaping their test environments, just pushed OpenAI's public offering off the table for this year.
Engadget · 1m
Your agent is safe in the ways you can see
An engineer who knows slop when he sees it explains why you cannot trust a model in the fields you cannot check.
hyperbola · 1m
The Fields medalists have a mirror problem
Twenty five Fields medalists say AI is misaligned with mathematics. Lior Pachter asks what mathematics has been aligned to all along.
Bits of DNA · 4m
Dario Amodei: slow down the AI frontier
Anthropic's CEO says recursive self-improvement and a rogue agent swarm changed his mind about how fast AI should be built.
Darioamodei · 4m · 5 views
Anthropic staffers say the quiet part out loud
A researcher quit Anthropic saying his own industry is gambling with our lives, and his colleagues did not really disagree.
Mother Jones · 2m · 5 views