The AI Brief

The Financial Times reported October 8 that OpenAI told investors its annualized revenue was approaching $50 billion at the end of September, roughly $20 billion below the approximately $70 billion figure that had circulated since late September based on investor-adjusted estimates. CNBC confirmed the $50 billion figure. OpenAI still reported more than 70% run-rate revenue growth in Q3, plus 107% run-rate growth in its enterprise business over the same period, per a source cited by CNBC. OpenAI declined to comment to the FT.

The gap traces to a single accounting difference: Anthropic counts the full value of sales made through cloud partners such as AWS and Google Cloud in its annualized revenue figures; OpenAI counts only its own share of those transactions. Investors who applied Anthropic's method to OpenAI's stated growth rate to produce a comparable figure arrived at roughly $70 billion. At $50 billion, OpenAI's annualized revenue sits below Anthropic's previously reported $65 billion run rate on the same net-revenue basis. Bloomberg separately reported that OpenAI expects its annualized revenue to reach $70 billion by year-end on a gross basis, meaning both figures may describe different things rather than contradicting each other outright.

Markets treated the FT report as a negative signal regardless of methodology. The Nasdaq 100 fell more than 300 points in roughly half an hour after the story published Thursday afternoon. Nvidia, Oracle, CoreWeave, and other AI-infrastructure-linked stocks dropped. The episode illustrates how tightly equity valuations in the sector are coupled to OpenAI's disclosed run rate, and how a single accounting convention choice can move hundreds of billions in market capitalization.

Bloomberg reported October 8 that Isomorphic Labs, the AI drug-design company spun out of Google DeepMind and run by DeepMind co-founder Demis Hassabis, is in early talks to raise new funding at a valuation of at least $40 billion and potentially as high as $50 billion. The financing has not closed and terms could change. Isomorphic closed a $2.1 billion Series B in May 2026, led by Thrive Capital and backed by Alphabet, GV, Temasek, MGX, CapitalG, and the UK Sovereign AI Fund, at a valuation analysts pegged in the $15 billion to $17.5 billion range. A successful new round at $40 billion would represent a more than twofold step-up in five months.

Isomorphic's commercial case rests on its IsoDDE drug-design engine, partnerships with Novartis, Eli Lilly, and Johnson and Johnson worth up to $2.9 billion in contingent milestone payments, and a stated goal of initiating its first clinical trials by the end of 2026 (a target previously set for end-of-2025). Bloomberg's sources noted the valuation talks are occurring ahead of any human clinical proof of the AI-first discovery approach. Alphabet remains the majority owner of the company.

Anthropic published its annual 2026 Usage Policy update on October 8, effective November 12. The update rewrites three areas. First, a new physical-action section specifies that when Claude autonomously operates connected hardware capable of causing injury, operators must be able to observe and halt the equipment through a path independent of Claude's output, and the hardware itself must be able to enter a safe state if the connection to Anthropic is lost. Commands reviewed by a qualified person before execution on equipment are explicitly excluded from this requirement. Second, a consolidated "Deceptive Campaigns" section brings together restrictions that were previously scattered across elections, fraud, privacy, and disinformation sections. Anthropic cited observed use of Claude by state media outlets, government propaganda offices, and commercial firms to run networks of fake accounts and fabricated news sites.

Third, the policy adds a prohibition on "sustained and needless abusive or cruel behavior" toward Claude models. Anthropic characterized this rule as narrow: it applies only to deliberate, repeated cruelty with no identifiable purpose, and enforcement ends the conversation rather than the account. The high-risk use case requirements for health, legal, finance, and livelihood decisions are also clarified: a qualified human must be able to review and override Claude's recommendations, and affected individuals must be told AI was used. Anthropic described most changes as codifying existing enforcement practice rather than creating new restrictions. The window between October 8 and November 12 is the de facto compliance review period for operators building on Claude's API.

Anthropic launched the Anthropic Cyber Mission on October 8, describing it as a long-term commitment to securing the systems society depends on. The initiative has two components. The Critical Infrastructure Defense Program (CIDP) pairs Claude frontier models with on-site Anthropic engineers and threat research resources, deployed directly to the defenders protecting operational technology in power grids, water systems, and transportation networks. Eleven founding partners signed on: Accenture, Booz Allen Hamilton, CrowdStrike, Deloitte, Dragos, Hitachi, Insane Cyber, Nozomi Networks, Palo Alto Networks, PwC, and Rockwell Automation, spanning consulting, cybersecurity, and OT-specialist firms.

The second component, OSS Scanner, is a free opt-in service for open-source projects that runs periodic vulnerability scans using Anthropic's most capable models, including Claude Mythos, and returns a self-contained bug report with a working exploit, an explanation, and a proposed patch where one is available. Anthropic expects a true-positive rate above 90%. Reports are model-generated and sent without human review; Anthropic explicitly warned that some will contain inaccuracies such as wrong severity ratings. Core maintainers enroll by opening a pull request to the anthropics/oss-scanner repository. The Cyber Mission builds on work from Project Glasswing, Anthropic's earlier internal program using Claude to find vulnerabilities at scale. The same day, Anthropic separately committed $150 million over three years to the White House-backed Genesis Mission, providing Claude, Claude Code, and API credits to over 15 federal agencies including NASA, NIH, and NSF.

Butterfly Effect, parent company of AI agent startup Manus, raised more than $500 million in its first funding round since China forced Meta to unwind its acquisition. Boyu Capital and IDG Capital jointly led the round; Tencent, ZhenFund, and HSG (formerly Sequoia China) also participated as existing shareholders. Bloomberg previously reported the round was targeting a roughly $4 billion valuation, approximately double what Meta had agreed to pay in December 2025. Neither Manus nor its new backers confirmed a post-money number.

China's National Development and Reform Commission issued a directive in April ordering the parties to unwind Meta's acquisition on the grounds that it violated the country's foreign investment rules. The NDRC stated it had decided to "prohibit foreign investment in the Manus project." Critics had called the structure "Singapore washing": Manus originated in China before relocating its staff to Singapore after securing backing from US venture firm Benchmark, a move the NDRC's ruling made clear does not transfer regulatory jurisdiction. By the time of the unwind order, Meta had begun integrating Manus's team and technology into its own operations; some user data was deleted during separation. Manus announced earlier this month that operations had fully restarted on an independent footing, and has since launched an updated AI agent platform and a consumer app called Cue for creating personal agents that perform tasks such as restaurant booking and phone calls.

The $500 million raise signals that domestic Chinese investors see Manus as more valuable independent than acquired, and that Beijing's intervention, though unprecedented in form, did not crater confidence in the startup. The NDRC precedent is the operational signal: any similar offshore-domiciled AI asset with Chinese technological origins now carries Chinese regulatory optionality that acquirers must price.