Sacks says Anthropic's safety talk is a business strategy. Strange strategy: they just walked away from a Pentagon contract over it.
Since January 2026, tech-industry figures and Trump administration officials have argued that Anthropic's AI-safety advocacy isn't sincere, that it's a strategy to capture regulators and box out rivals. Think Big PAC opened the attack on RAISE Act author Alex Bores on January 26. Pentagon leadership called Anthropic's CEO a liar with a "God-complex" in late February, after the company refused to loosen weapons-related guardrails for the military. Former White House AI adviser David Sacks kept the frame alive through June and July, and the All-In podcast closed the window on July 22 by accusing Anthropic of wanting to ban open-source AI.
The capture-strategy story runs into a problem fast: a company playing regulators for advantage doesn't usually eat a federal blacklisting, an executive order, and a Pentagon contract debarment rather than fold. Anthropic took all three rather than drop its weapons guardrails. That's a strange way to run a scam.
The mind-reading isn't the whole picture, though. Anthropic really has pushed state after state toward tougher rules, which tracks with safety as a selling point, and nobody has produced a receipt for the loudest claim, that the real plan is banning open-source AI. The pretext theory struggles. The positioning theory has something to stand on. Our score: mixed evidence.
Not one of the four predictions behind this claim came back clean. Two were mixed, two fell apart.
Being right about what happened and wrong about what it proves are different failures, so these two lines never get averaged into one score. How this works
The logic check
If the claim is right, these things should be true. Here's what we found. Each check says which version of the claim it's testing, the plain one most people would hear or the widest one the same words could carry, and whether it carries the claim's own burden or establishes the background it sits on.
If capture is the strategy, the regulatory outcomes of the safety camp's advocacy should advantage Anthropic and other incumbents over rivals.
MixedThe window's biggest regulatory actions actually hit Anthropic itself: a federal blacklisting, a push to purge its technology from government, and a preemption drive against its own state bills. That cost it government business rather than capturing regulators, though the state-bill burden on smaller teams is a real cost critics can point to.
If the safety fear is just pretext, the safety camp should drop its costly safety positions once they threaten revenue.
FalseAnthropic refused Pentagon demands to loosen surveillance and autonomous-weapons guardrails and accepted being cut off from federal contracts instead, the opposite of what you'd expect if the fear were faked.
If the real plan is banning open-source AI, there should be an actual proposal from Anthropic or safety groups to point to.
FalseThe claim traces back to Sacks reading "breadcrumbs" into a rival's essay, and his own follow-up post admits he isn't sure whether that essay was a confession or just a prediction. No ban proposal turned up anywhere.
If safety messaging is really competitive positioning, it should track the commercial and legislative opportunities it benefits from.
MixedReporting that Sacks himself shared shows Anthropic pushing states toward progressively tougher rules rather than one uniform standard, which fits the positioning read, and safety branding reportedly helped with recruiting. Still, the reporting describes a strategy, not proof of insincerity.
The receipts
Every quote, checked against its source.
David Sacks (former White House AI adviser)
"About 8 months ago, I warned that "Anthropic is running a sophisticated regulatory capture strategy based on fear-mongering." This take was controversial at the time; now look how many people are saying it."
David Sacks (former White House AI adviser)
"Woke lobotomized models are the enemy of American competitiveness."
Emil Michael (Under Secretary of War)
"It's a shame that [@DarioAmodei] is a liar and has a God-complex. He wants nothing more than to try to personally control the US Military and is ok putting our nation's safety at risk."
Pete Hegseth (Secretary of War)
"The Terms of Service of Anthropic's defective altruism will never outweigh the safety, the readiness, or the lives of American troops on the battlefield."
Sam Altman (CEO, OpenAI)
"Anthropic wants to control what people do with A.I."
Josh Vlasto (spokesman, Leading the Future / Think Big PAC)
"pushing legislation that would hand AI oversight to Albany bureaucrats to appease his ideologically motivated doomer supporters"
The full scoring
The raw numbers behind the verdict. What R/I/P/E/L and charge stages mean: how we score.
Research verdict: mixed evidence. This is the category the underlying research assigned before we graded the individual checks, and it came back the same on every card in this set.
R4 for recurring carriage by officials, Fox News, and X posts in the seven-figure impression range without saturating major outlets as their own voice; I4 for bad-faith default as the running lens plus essentialization and national-security stakes inflation, short of full purity policing; P4 for a recognizable current across at least four unrelated institutional nodes in the camp; E5 because the White House AI adviser, the Secretary and Under Secretary of War, the President, and a rival lab CEO voice it as their own; L4 because the safety camp answers directly with lawsuits, investigations, and counter-frames, with the tight two-way lock running through the mirror card.
Mirrors: The same NY-12 fight carries both frames in the same City & State article, with the PAC citing Bores's "AI Goliaths" attacks as proof of doomer opportunism while the safety side cites the PAC's spending as proof the industry is buying elections, and the July exchange repeats the pattern with Sacks reading Dean Ball's essay as a confession while the safety camp reads the blacklisting as retaliation.
Keeps coming back:
- NY-12 congressional race and the RAISE Act record (January)
- Anthropic's anti-ads Super Bowl campaign (February)
- Pentagon guardrail standoff, blacklisting, and executive order push (February to March)
- Kimi K3 leaderboard rise and guardrail-refusal posts (July)
- Politico state-regulation story and Dean Ball's essay (July)
Source record: counter-hate-monitor/counter-hate-work/special-reports/2026-tech-ai-fault-line/cards.json#pol-safety-fearmongers
The AI fault line, January to July 2026
Thirteen narratives, every carrier quote, the full scoring, and the coverage notes on what we couldn't verify. This page is the plain-language cut; that one is the whole thing, including the polarization methodology that defines every score here.