The Trump administration can blacklist AI firm Anthropic from all Pentagon contracts, a federal appeals court ruled Friday, in a decision that expands the executive branch’s authority to designate U.S. businesses as national security risks.
The D.C. Circuit Court of Appeals ruled 2-1 that the Defense Department had good reason to conclude that the company’s artificial intelligence products pose a supply chain risk to the department and its contractors because the company includes code in its models that restricts them from performing certain tasks. Anthropic has said that it will not allow its products to be used for autonomous weapons or mass domestic surveillance.
“The Department had ample support for its conclusion that the continued integration of Claude into the Department’s information systems, by the Department or its contractors, presented a statutorily covered national-security risk,” the court wrote. “As Anthropic admits, the company encodes restrictions into Claude that prevent the model from performing tasks that Anthropic wishes to prevent.”
You must log in or register to comment.
- 6 days
Let me get this straight… The judges agreed with Trump that Anthropic is a risk to the supply chain because… it cannot be used for autonomous weapons or mass domestic surveillance.
There’s no way the Supreme Court gonna be that stupid but if they do then a lot of U.S. employers are going to wake up one day and find themselves blacklisted.
- Zedstrian@sopuli.xyzEnglish6 days
The Supreme Court:
The constitution doesn’t mention any limits on AI-powered autonomous weapons or mass surveillance, so they clearly must be legal!
- kibblebits@quokk.auEnglish6 days
they’re pretty stupid… and this is project 2026, which ends with a full takeover of the USA.
So… I put my money on them doing whatever looks like handmaids tale.
If national security can depend on autonomous decision making and this software refuses it, isn’t that a risk to national security if it’s used?
What if a supplier uses it without knowledge, and somewhere in the kill command chain it rejects it?
I think it’s pretty stupid and board because this also impacts things like a contractor writing an email, but the logic that something will refuse a command that can be considered national security seems reasonable?
This also doesn’t mean we should be doing this, but if we are, then yes it’d be a risk.
- 5 days
But can they label all AI as a national security risk? Anthropic’s only additional crime was not wanting to be involved in weapons development or mass surveillance.
Let’s be clear, not wishing to do physical harm, that is not inherently a security risk; whereas the inverse, AI that will play ball with the Pentagon, and then relying on that AI to pull triggers, that’s an extreme security risk.
I mean no surprise that trump would get it backward.