Anthropic, the AI company behind the Claude chatbot, may have created a regulatory problem for itself simply by talking too much about the dangers of its own technology.
According to a Financial Times analysis by Clara Murray, Anthropic's 2026 official statements and posts used AI risk-related terms roughly eight times more often than rival OpenAI did. The FT analysis found that the company warned about the dangers of advanced AI far more frequently than OpenAI over the course of this year.
The concern flagged in the analysis is blunt: by repeatedly and publicly emphasizing how powerful and potentially hazardous its systems are, Anthropic could hand regulators a ready-made argument for restricting where and how its technology can be sold abroad. In other words, the company's own safety messaging might become evidence in the case for an export ban.
Separately, a report carried by Yahoo Tech notes that a crypto executive has disputed claims that an Anthropic system referred to as "Mythos" breached NSA systems. The executive's pushback suggests those breach allegations are contested rather than settled, though the available material does not detail the original claims or who first made them.
Taken together, the two items point to a company navigating heightened scrutiny over both its security posture and its public framing of AI risk.
Why it matters: how an AI company describes the risks of its own products is no longer just a branding choice — it can shape whether governments let that technology cross borders.