Anthropic Tests AI Agent Deal-Making Internally and Partners With Law Firm Freshfields

Anthropic reportedly ran a closed, 1-week internal experiment in its San Francisco office creating a classified marketplace where AI models negotiated and conducted deals on employees' behalf, though the company has not published transaction counts or dollar values. The test sits alongside an April 23, 2026 collaboration agreement with global law firm Freshfields—which employs more than 2,800 attorneys—to jointly develop AI tools for legal services without disclosing financial terms. Freshfields is working directly with Anthropic's legal and product teams on the Cowork agentic platform across due diligence, contract review, and document drafting, receiving early access to future models while maintaining human attorney sign-off. The push, referred to by industry outlet Artificial Lawyer as "Project Deal," arrives as Anthropic reportedly pitches investors on a $110B annualized run rate (with Q3 revenue pegged at $16.8B in separate reporting) and navigates an ongoing FTC inquiry into AI agent practices alongside earlier AI safety legislative proposals. "Partnering with Anthropic strengthens our ability to co-innovate at pace and to bring new capabilities into our work in a way that is secure, compliant and focused on client needs." — Gil Perez, Chief Innovation Officer at Freshfields.