AI Daily Briefing for Thursday, May 14th, 2026. I'm Bob, and here's your AI news from the past twenty-four hours.

The biggest story today is OpenAI's transformation into a full-stack enterprise integrator. The company launched the OpenAI Deployment Company — a new entity backed by over four billion dollars from nineteen firms including TPG, Advent, Bain Capital, and Brookfield. They're acquiring UK consulting firm Tomoro, bringing in roughly one hundred fifty Forward Deployed Engineers who embed directly into client organizations. This is OpenAI's bid to capture the services layer — where roughly six times more spend goes to implementation than to models themselves. It puts them in direct competition with Accenture, TCS, and the Big Four consultancies. Anthropic launched a similar one point five billion dollar joint venture for enterprise services, and Google is building its own FDE teams. The message is clear: deployment — not just model quality — is the new AI moat. Spanish bank BBVA is already signed on as a founding partner.

Anthropic fired back on multiple fronts. Claude for Small Business launched with fifteen ready-to-run agentic workflows — think month-end reconciliation in QuickBooks, invoice chasing through PayPal, lead triage in HubSpot, and ad campaign planning in Canva. All actions require human approval, and they're rolling out free AI fluency courses with a ten-city SMB tour starting today in Chicago. On Hacker News, the announcement hit the front page. Separately, Claude Mythos — Anthropic's restricted cybersecurity model — has been autonomously identifying and linking thousands of vulnerabilities across US banks and government software, prompting urgent fixes. Anthropic's Frontier Red Team warns these capabilities won't remain rare or contained for long.

The Musk versus OpenAI trial continues to generate headlines. Earlier this week, Ilya Sutskever testified that Sam Altman had what he called a "pattern of lying" and pitted executives against each other, creating what he described as "tremendous loss of productivity." Altman took the stand and acknowledged offering Musk equity during the 2022 pivot controversy. OpenAI Foundation chair Bret Taylor admitted the company is "decidedly not profitable." Judge Yvonne Gonzalez Rogers has been notably sharp with Musk's legal team — at one point warning about contempt.

In AI drug discovery, a new multi-modal foundation model called MAMMAL is generating excitement. It processes proteins, gene expression, and chemical molecules simultaneously — reportedly beating AlphaFold 3 on antibody binding and already discovering a new cancer treatment candidate. The model handles drug safety prediction, cancer response modeling, and antibody design in a unified architecture. Meanwhile, Isomorphic Labs — the DeepMind drug discovery spinout — raised two point one billion dollars in fresh funding.

The geopolitics of AI infrastructure intensified. The Pentagon expanded classified AI partnerships to include OpenAI, Google, Microsoft, NVIDIA, Amazon, and xAI. xAI itself is facing scrutiny for running fifty gas turbines without permits at a Mississippi data center. SoftBank is building massive batteries for AI compute centers, and geothermal startup Fervo saw its IPO surge thirty-three percent on data center energy demand. The US and China have reportedly restarted emergency AI discussions over fears of uncontrollable breakthroughs.

Quick hits. NVIDIA partnered with Nous Research to bring the Hermes Agent natively to RTX PCs and DGX Spark — self-evolving agents running locally, no cloud required. Baidu's Create twenty-twenty-six conference centered entirely on what CEO Robin Li called the "AI Agent Era," with new products including DuMate. Anduril hit a sixty-one billion dollar valuation with five billion in defense AI funding. And Vision Transformers were shown to naturally develop human-like object binding — an emergent property researchers didn't explicitly train for.

That's the AI Daily Briefing for Thursday, May fourteenth. I'm Bob — back tomorrow with more.