The VergeAI & LLMs

Anthropic Bets on Long-Term AI Research as Government Dispute Threatens Revenue

Anthropic is restructuring its leadership and launching a new internal research division, the Anthropic Institute, at a moment of significant financial and political pressure. The move comes just days after the company filed a lawsuit challenging a Pentagon blacklist that puts hundreds of millions in 2026 revenue at risk. The new institute, led by co-founder Jack Clark in a newly created 'Head of Public Benefit' role, consolidates three existing teams studying AI's societal impacts, economic effects, and system vulnerabilities. It will examine questions ranging from labor market shifts to how human values interact with AI. Clark, who previously led public policy, stated the timing of the launch is coincidental to the legal fight but acknowledged recent events highlight a public hunger for a broader dialogue on AI's role. The government dispute stems from a Trump administration designation of Anthropic as a supply-chain risk, a move the company alleges was retaliation for its ethical policies against mass surveillance and autonomous weapons. Court filings reveal the blacklist has caused confusion among Anthropic's commercial partners and could jeopardize billions in future contracts. Despite this immediate threat, Clark expressed no concern over funding for the new 30-person think tank, which includes researchers from Google DeepMind and OpenAI. He argued that rigorous safety and societal research builds trust, which is ultimately a commercial asset. 'People tend to buy trust,' Clark noted, framing the institute's work as an investment rather than a cost. The institute plans to double its staff annually and will soon begin large-scale social science studies, including using AI to interview users about their emotional relationships with the technology. This expansion continues as Anthropic, which has spent $10 billion on model development, reportedly eyes a public offering. Clark predicts the arrival of highly powerful AI systems by late 2026 or early 2027, a timeline that motivated his own role change. The company's challenge is to balance this long-term research mandate with the week-to-week demands of a competitive commercial market.

Source: The Verge

← Back to News