- Anthropic is investigating unauthorized access to Claude Mythos, a restricted cybersecurity model offered only to vetted enterprises, cleared organizations, and select government agencies.
- Worth monitoring as a precedent for tiered-access frontier-model security incidents.
- Sources scanned: TechCrunch AI, VentureBeat AI, The Decoder, Bloomberg, CNBC, Techmeme, Invezz, Axios, Import AI, TechXplore, The AI Track, llm-stats aggregator (covering OpenAI, Anthropic, Google/DeepMind, Microsoft, Meta, Amazon, Nvidia, DeepSeek, Adobe, plus Harvard Medical School / Beth Israel and arXiv).
Snapshot — April 21, 2026
5 stories
Google DeepMind released Gemini 2.5 Ultra with a 2M-token context window, native multimodal tool use, and an LMSYS Chatbot Arena Elo of roughly 1,421 — the highest publicly measured score to date. The launch pairs with a newly formed DeepMind coding team explicitly positioned to rival Anthropic's Claude Code franchise.
The European Parliament voted 569 in favor of a Digital Omnibus proposal to delay full enforcement of high-risk AI obligations from August 2, 2026 to December 2, 2027. The Council approved its negotiating mandate on March 13; trilogue negotiations are ongoing, but prohibited-practice and GPAI obligations remain enforceable today.
Anthropic has reportedly reached roughly $30B in ARR versus OpenAI's $25B, capping 30x growth in 15 months. The surge is credited to Claude Opus 4.7 (released April 16), which now leads most public benchmarks and is live across Claude.ai, the API, AWS Bedrock, Google Vertex AI, and Microsoft Foundry.
Meta unveiled a $600B AI investment plan anchored by its new Muse Spark model, positioned as a driver of productivity and workforce transformation across the U.S. economy. The scale of the commitment escalates the hyperscaler capex arms race already underway among Microsoft, Google, and Amazon.