🧠 Food for thought

Implications, context, and why it matters.

The government’s ‘lawful use’ clause stems from a specific conflict over Anthropic’s Pentagon dispute

• The “any lawful use” or “all lawful purposes” wording grew out of a recent Pentagon clash with Anthropic over limits on how its Claude model could be used 1.
• Anthropic sought contract terms that would clearly block Claude from mass surveillance of Americans or from fully autonomous weapons that target people without human input 1.
• The US Defense Department said those activities already face legal limits plus internal Defense Department policies, so it insisted on the option to use Claude for “all lawful purposes” without vendor restrictions 1.
• The standoff prompted President Donald Trump to order federal agencies to “immediately cease” using Anthropic’s technology, while the Defense Department got up to six months to phase it out; OpenAI soon said it had reached an agreement with the Pentagon 1.

AI safety principles are bending under commercial and political pressure

• The episode shows how business and political forces can strain the AI industry’s public safety promises 2.
• As competition tightened, OpenAI removed explicit bans on military and warfare uses from its policies in 2024 2.
• Anthropic also updated its Responsible Scaling Policy (its internal framework for deciding when to slow or pause the rollout of more capable models) and said it would not on its own delay a new model only because it still lacked a way to make the model safe 2.
• The money at stake is large, with OpenAI’s annualized revenue at $25 billion and Anthropic’s recently topping $19 billion, which shrinks the gap between the companies 2.