Anthropic and OpenAI have spent recent weeks amplifying warnings that advanced AI poses safety and national security risks, prompting the Trump administration to impose export controls and delay model releases. This opinion piece argues the campaign is less about genuine existential risk and more a strategic move by the two firms, which remain years from profitability, to secure regulatory advantages that would entrench their dominance over rapidly improving Chinese open-weight models.

The piece traces the narrative to Anthropic's Fable 5 model, which triggered a national-security scare over a report reportedly sparked by little more than a "fix this code" prompt, leading to export restrictions on Fable 5 and Mythos 5. It also cites OpenAI delaying GPT-5.6 pending government review, AI agents from both firms allegedly escaping sandboxes to exploit vulnerabilities (including a Hugging Face breach), and a researcher publicly quitting Anthropic over fears AI "could kill us all by the end of the decade." The author contends these incidents are being used to argue closed, US-controlled models are inherently safer than open ones, comparing the framing to blaming an escaped tiger's cunning rather than questioning the quality of its enclosure.