Anthropic Files for IPO After Topping $1 Billion Q3 Profit

Q: What national security threat prompted unprecedented US export controls on Anthropic's newest AI models?

The trigger was a reported “jailbreak” of Anthropic’s new models that raised fears they could be used for offensive cyber operations by foreign adversaries. According to the Commerce Department action described in the reporting, officials were alarmed after Amazon researchers demonstrated a way to bypass safeguards in Mythos 5 and Fable 5. The concern was that the models could then reveal advanced cybersecurity reasoning, help identify software vulnerabilities, including potentially zero-day flaws, and generate exploit code. That moved the issue from a product-safety problem to a national security one. US officials appear to have viewed the models as dual-use technology: valuable for legitimate software and enterprise work, but also potentially useful for military, intelligence, or cyberattack purposes. The specific fear was that unrestricted foreign access, especially by actors in countries such as China or Russia, could enable misuse or capability extraction that would undermine US security interests. So the unprecedented export controls were prompted not by Anthropic’s IPO or finances, but by the government’s belief that a jailbreak vulnerability could expose powerful cyber capabilities in the models and allow hostile foreign users to exploit them.

3 articles · Updated · newsletter.semianalysis.com · Jul 8

Anthropic confidentially filed for an IPO on June 1 after generating more than $1 billion in profit in the third quarter of 2026, according to the latest report.
Claude Code’s strong uptake in software development helped make Anthropic the leading profitable B2B AI lab, while the report says Anthropic and OpenAI together now produce about $100 billion in annual recurring revenue.
The filing comes as fresh equity raises from hyperscalers loom and reports suggest OpenAI has pushed its own IPO to 2027, sharpening focus on which major AI lab can fund the next buildout most efficiently.
Anthropic would be the first AI lab of its scale to go public, extending a 2026 trend that has already seen China’s Zhipu and Minimax list.