Too long. Need 60-90. Let's shorten: "Pentagon Gives Anthropic Ultimatum as DeepSeek

Anthropic’s refusal to grant the Pentagon unrestricted Claude access has pushed the standoff into a legal‑policy showdown, with Defense Secretary Pete Hegseth warning that the company could be labeled a supply‑chain risk and subjected to the Defense Production Act if it does not comply by Friday, according to Reuters. The ultimatum follows a series of coordinated “distillation attacks” that Anthropic says were carried out by DeepSeek, Moonshot (Kimi) and MiniMax, which together generated more than 16 million illicit queries and 24 000 fraudulent accounts to siphon Claude’s capabilities, as detailed in a report by Siddhesh Surve.

In response, Anthropic announced a comprehensive rewrite of its safety policy, dubbed “RSP 3.0,” and the launch of an enterprise‑plugin ecosystem aimed at bolstering control over its models. The move was highlighted in a Zecheng Intel Daily briefing, which noted that the company rolled out the new safety framework and plugins on the same day it faced the Pentagon’s demand, positioning the changes as a defensive measure against external exploitation.

The Pentagon’s pressure stems from Anthropic’s unique status as the sole AI firm whose models run on classified military systems, a fact underscored by multiple outlets including The Verge. If the Department of Defense invokes the Defense Production Act, it could compel Anthropic to prioritize government workloads and potentially restructure its operations to meet unrestricted access requirements, a scenario the company has publicly resisted, as reported by TechCrunch.

Analysts warn that the dispute could reverberate across the broader AI market, where the recent piracy episode has already raised concerns about model security and intellectual‑property theft. Anthropic’s aggressive stance—refusing to bend to the Pentagon’s deadline while tightening internal safeguards—signals a willingness to challenge governmental leverage, a position that may set precedents for how private AI firms negotiate national‑security demands in the future.