Grok is a burning toxic wate dump. Their quality control is awful. It's the alpha Romeo of AI
It's a massive PR coup for Anthropic.
They already dominate the safety first and ethical AI markets and are growing three times faster than their main rivals.
Their willingness to stand up to bullies will only enhance their reputation within more sober business sectors.
Anthropic AI ddn't agree
I asked anthropic how it would recover from being classed supply chain risk..
The real business risk: The bigger threat isn't the $200 million contract loss — it's that Anthropic's large enterprise customers who also hold Pentagon contracts may feel pressure to drop Claude from their workflows entirely.
It also agreed that it was likely a negotiation tactic:
The catch: Even if it's a tactic, it has real teeth. Anthropic is preparing for a potential IPO with a $380 billion valuation, and the designation could spook enterprise customers and investors regardless of its legal merit.
CNN So the leverage is real, even if the framing is theater.
There's an inherent awkwardness in Anthropic's position:
The "but we have limits" argument is slippery. Anthropic was apparently
willing to work with the Pentagon — just not on mass domestic surveillance or fully autonomous weapons. But once you're in that relationship, where exactly does the line hold? Defense contracts have a way of expanding in scope, and "we'll help with some war stuff but not the bad war stuff" is a hard position to defend publicly or legally.
The public may not make fine distinctions. To a lot of people, "Department of War" + "AI" = bad, full stop. The nuance of "we refused
this specific use case" may not land — especially if Anthropic later signs a narrower deal, which is probably the likely outcome here.
It could look like price negotiation dressed up as ethics. Which, as we discussed, it arguably is — from
both sides. The Pentagon is using the designation as leverage, but Anthropic is also using its public stance as leverage to negotiate better terms. Neither party is purely principled here.
seems quite smart
www.cnbc.com
