Anthropic Deputy Director of Information Security Jason Clinton issued a stark warning to security chiefs: the worst is yet to come for the industry, and the industry has roughly 18 months to withstand the growing wave of threats.
Clinton urged top management to go to boards of directors immediately for resources—and to ask not for the deployment of AI tools, but for basic security measures. He characterized the last four days as difficult, noting that the situation will get significantly worse before it begins to improve.
Why the count is in months, not years
Clinton's assessment is part of a broader picture of the threats network users have faced recently, including risks associated with artificial intelligence. Back in August 2026, he warned that in a few months a "tsunami of vulnerabilities" would hit infrastructure, and there would be catastrophically insufficient time to install patches in critical systems.
As containment measures, the expert proposed two approaches right now:
- air gap — physically disconnecting systems from external networks;
- data diode — a device that allows information to pass in only one direction and does not allow an attacker to transmit anything back.
The reason for the first post was a warning from the US National Security Agency. In August, the NSA, together with other agencies, issued guidance on protecting programmable logic controllers—devices that control industrial equipment. The document directly spoke of attacks on Siemens S7 series controllers by attackers using AI.
Criticism in response
Not everyone received the message positively. Some industry specialists objected: when experts of such a level start publishing posts like this, it is already too late to act, and what is happening resembles the opening credits of a disaster movie.
Clinton himself, however, insists otherwise. By his logic, right now, while the crisis is in plain view, security services have a chance to obtain the resources they usually lack. Also worthy of separate attention is Anthropic's statistics: more than 67% of the company's blocked accounts were used for cyberattacks using AI.
Cryptalist analysis: Anthropic's warning reflects a fundamental shift in the threat landscape: generative AI radically lowers the barrier to entry for attackers and accelerates the vulnerability exploitation cycle, especially in industrial and infrastructure systems. For the crypto sector, this is a signal of heightened risk for custodial services, bridges, and smart contract infrastructure, where the compromise of keys or logic controllers is fraught with direct financial losses. Investors and platform operators should factor heightened cyber risks into their assessment of project resilience over the next year and a half.