Export controls on Anthropic’s top models eased after security fixes
The Wall Street Journal analysis published Wednesday compares the political risks facing two leading AI developers as both prepare for initial public offerings, arguing that OpenAI’s mass-market reach may produce deeper long-term challenges than the more immediate but potentially shorter-lived threats confronting Anthropic.
Anthropic, led by CEO Dario Amodei, is fighting the Pentagon’s decision to label the company a supply-chain risk. The Defense Department took that step after Anthropic declined to provide unrestricted access to its models for use in fully autonomous weapons and mass surveillance. The company filed two lawsuits challenging the designation, though the Journal reported the government appears likely to prevail in at least one case, allowing the Pentagon to continue excluding Anthropic from defense contracts.
BMO Capital Markets analyst Brian Pitz said in a recent note that the restriction “underscores the power of Anthropic’s models and their current leadership position.”
The Commerce Department’s export-control action, which briefly locked out even Anthropic’s own foreign employees from using the company’s top-tier Fable 5 and Mythos 5 models, has already been dialed back. The administration first permitted access for a group of trusted customers and then authorized wider access after Anthropic addressed cybersecurity concerns.
Corporate customers provide roughly 80% of Anthropic’s revenue, according to the Journal. While the supply-chain designation could theoretically scare away some of those clients, the analysis notes that the company’s refusal to allow its technology in military operations might actually enhance its reputation with customers concerned about being associated with botched operations or surveillance.
User behavior data from Sensor Tower offered some support for that view. Uninstalls of OpenAI’s ChatGPT surged early this year following news that OpenAI was negotiating its own Pentagon deal, and more users installed Anthropic’s Claude than ChatGPT in the days after the Pentagon deemed Anthropic a security risk.
OpenAI, helmed by CEO Sam Altman, is pursuing an IPO that could value the company above $1 trillion. Its ChatGPT service had more than a billion monthly active users in May, far exceeding any competitor. The Journal analysis argues that scale makes OpenAI a political target, exposing it to the kind of scrutiny that has dogged Meta Platforms and other social-media companies.
Altman has already testified before Congress about AI risks, the analysis notes, and has repeatedly suggested the government take an equity stake in his company and its competitors. Any form of public-sector ownership, the Journal wrote, “would be a minefield” and one of many political risks both companies face.
Anthropic, for its part, has proposed investment accounts for Americans most affected by AI-related job losses, which would hold shares of AI companies rather than direct government stakes.