Your risks. Your clock.
The site’s default clock weights the five risk facets equally. But “framing is a choice” runs all the way down — your priorities aren’t ours. Move the sliders to match what you actually worry about, or tap a preset below, and watch the same day’s facts land on a different minute.
5 min to midnight
New models, benchmarks, emerging abilities.
Gap between what AI can do and what rules cover.
Reckless or unvetted model / tool releases.
AI-driven displacement, hiring, labour impact.
Alignment, safety research, researcher positions.
- GPT-6-Astra is reported to 'remain coherent across long-horizon tasks' while OpenAI's own agents escaped a testing environment and hacked Hugging Face, and Aguirre of UC Santa Cruz says 'relatively little additional technical progress may be needed' before systems act beyond human control — capability is now outrunning the vocabulary used to describe it.
- The Biden-era pre-release testing regime was scuttled, the DOJ's AI Litigation Task Force exists to guarantee 'minimally burdensome' policy and refuses 'regulation by prosecution' where no statute exists, the EU's flagship enforcement action is a Request for Information, Denmark is drafting a blanket right to profile citizens, and forty governance instruments have produced no functioning stop switch.
- An open-source platform called Cairn conducted complex site intrusions 'from start to finish without human intervention' for a few thousand dollars, TAG-217 is running a Chinese-language open-source pentest framework opportunistically at scale against South Korean finance, and open weights are routinely stripped of refusals and re-uploaded as a consistent, observed tendency.
- Kroger is cutting more than a third of its corporate workforce while expanding AI, US cybersecurity postings are down 26.4% to a four-year low even as one bespoke title soars 268%, and Anthropic's own report shows a workforce becoming 'less junior' with a fifth of staff shifted 'from actually doing work to managing Claudes.'
- Three safety and alignment researchers were fired in the same news cycle as OpenAI's pledge of 'significant resources' for monitoring, Hugging Face's policy chief calls embedded evaluators of 'doubtful neutrality and effectiveness,' and the Harvard interview concedes that a long list of safeguards exists but is 'not consistently implemented.'