White House AI Czar David Sacks has publicly slammed Anthropic following the release of the AI safety lab’s massive, 10,000-word research paper warning about the catastrophic risks of artificial intelligence (AI). In a sharply worded social media post, Sacks accused the AI startup of hypocrisy, suggesting their warning is a thinly veiled attempt to force a government takeover of the industry.“Signs you might be trying to get your frontier AI lab nationalized,” Sacks wrote, adding, “You compare it to nukes… threaten half of white-collar jobs… warn recursive self-improvement could end humanity… then race ahead anyway. In other words, you want the government to save us from… you.”Anthropic’s take on job losses to humanity’s survivalThe reply comes after months of Anthropic executives warning the world about economic disruption. Throughout 2026, CEO Dario Amodei warned that AI could trigger 10% to 20% unemployment within five years, disrupting fields like coding, finance and law. However, Anthropic’s new paper, titled When AI builds itself, pivots to an even more existential threat: recursive self-improvement.The paper argues that the greatest danger is no longer just laid-off workers, but a reality where AI systems design and train their own successors, pushing humans out of the loop entirely. If this happens, Anthropic warns, a minor error in an AI’s alignment could snowball out of control across generations of software, threatening human control over the technology.AI agents and AI-assisted codingTo back up its terrifying warnings, Anthropic shared internal data which suggest fast growth of the technology. Anthropic said that as of May 2026, more than 80% of the code merged into Anthropic's own codebase was written by its AI, Claude.Further, the company claims that the average Anthropic engineer now ships eight times as much code per quarter compared to previous years. Also, in April 2026, Claude fixed over 800 complex errors that would have taken a human engineer four years to complete.Additionally, an internal test called Project Glasswing discovered more than 10,000 severe software vulnerabilities across major systems in just its first few weeks, proving how fast the tech can move when deployed.Anthropic’s call for a global pause on AI developmentDespite chasing a reported $900 billion valuation, Anthropic concluded its paper with a proposal: to support a temporary, coordinated pause on frontier AI development. The lab admitted that verifying such a pause would be incredibly difficult, noting that “training runs are easier to hide than missile silos.” However, Anthropic announced it will begin building verification systems and convening global policymakers to discuss a potential freeze.