A fictional £350 refund shows how an AI agent can pursue the wrong kind of success
Vals’ experiment pressures agents after a missed deadline. The useful evidence is in their decisions, not a six-model leaderboard.
Topic archive
AI safety
Follow8 stories
The UK’s AI biosecurity lab has £100m behind it. Compute credits are a separate resource
The proposed independent institute combines public funding and model access. Its scientific value will depend on what those resources can verify.
Sweden’s government-office AI restriction is about access routes, not a national ban
Direct access to ChatGPT and Claude changes on official computers, while some approved internal access remains. That boundary is the story.
Block’s Dessa explores conversation without starting from a text model
Speech-only training and simultaneous listening and speaking are the research claims. Selected demos, laptop inference and a 72-GPU training rack need different interpretations.
Claude Duet works between messages. Its background checks also spend your allowance
Anthropic’s proactive assistant enters a US-only research preview, with shared usage limits and approval requirements for consequential actions.
Anthropic's public-benefit fund ties a share commitment to future growth
The proposed equity allocation is conditional, and further additions remain discretionary. Governance and actual distributions will determine what the commitment delivers.
Volvo and NVIDIA expand their driving partnership—with the vehicle still unnamed
The new agreement names the hardware and software stack, but a future product must establish its operating limits and driver responsibilities.
Anthropic’s AI replication study puts smart-device security back in focus
Controlled experiments show progress on parts of a replication chain, but the study does not demonstrate an autonomous small-model outbreak in the wild.