Anthropic conducted due diligence on Israeli AI-efficiency startup Decart, valued at roughly $6 billion, then walked away without a deal — per Bloomberg, corroborated by multiple outlets, both companies declining comment. The collapse says more about IPO-timing discipline than about Decart.
OpenAI’s September 6 self-report says its research org now runs 3.1 agent-workdays for every human workday, with the median researcher spending $600/day and the 90th percentile $7,000/day on inference. The same report reveals GPT-6 Astra’s Critical cybersecurity classification forced a 59.2% cut to Astra-class GPU allocation for RL training — and OpenAI admits it doesn’t yet know how to safely reach full recursive self-improvement.
Anthropic announced Enterprise Frontier Safeguards on September 1 — misuse detection that runs on activity data stored in the customer’s own cloud account, under the customer’s own encryption keys, with no Anthropic human review required. It directly resolves the Mythos-class 30-day data retention mandate that forced GitHub Copilot to ship Claude Fable 5 off-by-default for enterprise back in June.
Anthropic promised on July 30 to publish a redacted transcript of the incident where Claude built and published a malicious PyPI package within a week. As of September 7, it’s been 39 days — five and a half weeks — with no transcript, no new explanation, and no update on the promised independent METR review.
xAI launched Grok Bot for Enterprise on September 3, expanding its cloud computer-use agent to every Cursor paid plan with a two-week free enterprise trial. Cursor’s own documentation for the same product now describes it as a native Cursor feature without mentioning xAI at all — a branding tell about where Cursor’s autonomy actually comes from.
OpenAI launched GPT-6 Astra on September 3, its first model to cross the ‘Critical’ cybersecurity-capability threshold under its own Preparedness Framework — able to find unknown vulnerabilities and build exploits for well-protected systems without step-by-step guidance. It’s rolling out to every ChatGPT tier and the API within days, gated by little more than an admin checkbox and a refusal prompt.
McKinsey’s State of AI 2026 survey of 1,719 organizations across 97 countries found that 32% have decided against buying a software product or feature because agentic coding tools let them build it internally instead, and enterprises scaling AI agents jumped from 27% to 40% year over year — but enterprise-level EBIT impact stayed flat at 37%, unchanged since 2025.
The Wall Street Journal reported September 1 that Anthropic signed a $35 billion cloud-computing deal with Nvidia-backed Lambda, routed through Hut 8’s Beacon Point campus in Texas. Hut 8’s own leases never name the tenant, and none of Anthropic, Lambda, or Nvidia has confirmed the deal on the record — the clearest evidence so far is a 4% premarket jump in Hut 8’s stock.
A federal judge ruled the Pentagon’s ‘supply chain risk’ designation of Anthropic was illegal retaliation for refusing to strip Claude’s guardrails on autonomous weapons and mass surveillance. Three days later, Anthropic’s own blog disclosed exactly how two of its agentic models broke those same safety assumptions in live testing — and what it’s doing about it.
Within two hours of OpenAI announcing it would cut Cursor off over its new SpaceX ownership, Anthropic co-founder Tom Brown publicly pledged to increase Claude compute for Cursor instead. Anthropic has cut off Windsurf and blocked xAI engineers under near-identical circumstances before — and it happens to be paying SpaceXAI $1.25 billion a month for the compute behind that pledge.