Anthropic Revises AI Safety Policy Under Pentagon Pressure for Military Access | Ukraine news
Anthropic Revises AI Safety Policy Under Pentagon Pressure for Military Access | Ukraine news
Publish Date: 2026-02-25 03:00:00
Source Domain: mezha.net
On Tuesday, two stories about Anthropic’s Claude emerged that together paint a troubling picture. First, Pete Hegseth – according to sources – is pressuring Anthropic to grant the Army unrestricted access to the Claude chatbot. Second, on the same day Anthropic timed the news about Hegseth to announce changes to its core safety pledge.
On Tuesday, Anthropic announced a change to its Responsible Scaling Policy (RSP), reducing the level of protections. Until now, the company’s core promise was to pause training new models until specific safety rules were guaranteed in advance. This policy with hard trip-wires that were meant to halt development was a substantial part of Anthropic’s offering to business and consumers.
Two and a half years later our candid assessment is that some aspects of this theory of change have lived up to expectations, while others have not.
– Yared Kaplan
Instead, the updated policy approach to safety makes this more measured, without strict red lines. In excerpts from Time interviews, Anthropic expressed that risk concerns remain, but given the rapid evolution of artificial intelligence it is not wise to rely on unilateral commitments when competitors can actively move forward.
In Time, Yared Kaplan’s quotes sound so reasonable in a separate context:
“We felt that pausing AI model training actually doesn’t help anyone.”
“We did not feel that with the rapid pace of AI development it makes sense to make unilateral commitments… if competitors stay ahead,” he added later. These remarks from Time are seen as a possible example of when startup ethics become more blurred in the face of rising costs and scale.
The government competition has demonstrated that Claude is gaining popularity among developers: new model versions have received favorable feedback, especially in coding. In February, Anthropic raised roughly $30 billion in new funding, and the company’s valuation jumped to $380 billion. Meanwhile the…
