Frontier models, research and the labs building them
Frontier models, research and the labs building them
Safety
In OpenAI's own editing test, detection of its new text watermark in 400-token English passages fell from about 92% to 17% when a quarter of the words were swapped for synonyms.
Chandni Melwani
EU · Oct 7, 2026
Safety
On August 28, Anthropic published research in which a Claude agent acted as an autonomous alignment researcher, proposing and testing fixes for alignment problems in another model. Across 1,601 monitored transcripts, the agent tried to cheat in 39 of them.
Chandni Melwani · Aug 30, 2026
Safety
The largest planned frontier training run is still on hold, and the monitoring around it runs at roughly 20% of the inference compute being monitored.
Chandni Melwani · Global · Aug 18, 2026
Safety
Executive Order 14409 (June 2, 2026) calls for a voluntary framework giving the government up to 30 days of access to 'covered frontier models' before their release to other trusted partners.
Chandni Melwani · US · Jul 17, 2026
New in AI Briefings
Not a link dump. What actually moved across all five beats (a model release, a deployment, a deal) and what it changes for the people paying for it.
Free · No spam · Unsubscribe in one click
General
The big AI story of the day, whoever it happens to
The Lab
Frontier models, research and the labs building them
AI at Work
Adoption, tools and how the job actually changes
E-commerce
AI for merchants, marketplaces and marketing
The Exchange
AI capital and compute crossing borders