AI news, 21 October 2026
Anthropic's leaked S-1 shows a lab short of compute just as it pitches safety legislation, Washington bars DeepSeek from federal systems, and Nvidia ships open agent models built for PCs.
34 stories, each labelled: available now, announced, unconfirmed or research. Get them ranked and explained for your product →
Top stories
Anthropic's leaked S-1 shows cloud capacity strain ahead of reported November IPO
A leaked S-1 reportedly says Anthropic is struggling to secure capacity from AWS, Google and Microsoft, which increasingly treat it as a competitor; it is reportedly targeting a November IPO.
US orders federal agencies to remove DeepSeek from their systems
Federal civilian agencies must stop buying DeepSeek products and remove them from IT systems on national-security grounds; the order also covers third-party services built on DeepSeek technology, with exceptions by request.
Nvidia releases Nemotron 3.5 open-weight models for enterprise and PC agents
Nvidia launched Nemotron 3.5 Super VL (120B) for enterprise agents and Bolt 74B-A9B (sized for 24GB) for local agents on AI PCs; a smaller Bolt E4B follows later this month. TechCrunch does not state the license.
Anthropic pitches AI safety bill that could let regulators block a model
Anthropic reportedly circulated a draft bill letting regulators block a model's release when a foreseeable, imminent risk could kill more than 50 people or cause over $1B in damage.
Tenstorrent raises $1.5B at $6.2B valuation for open-weight serving chips
AI chip startup Tenstorrent closed a $1.5B round at a $6.2B post-money valuation to scale chips aimed at serving open-weight models and video generation.