100+ cybersecurity experts: lifting Anthropic export controls would help, not hurt, US security

The paradox: loosening controls could strengthen security

An interesting open letter dropped this week. Over 100 cybersecurity experts publicly urged the US government to lift export controls on Anthropic's latest AI models. Usually "AI export controls" tighten in the name of security — here the argument flips: loosening them would actually help.

What happened

The backdrop is recent export restrictions on Anthropic models like Fable 5 and Mythos 5, aimed at blocking foreign access. Experts see a side effect: it also blocks allied security teams from using Claude defensively. Amazon CEO Andy Jassy had earlier raised similar concerns with senior Trump administration officials.

Why it matters

AI is a weapon for attackers but also a core defensive tool — intrusion detection, vulnerability analysis, threat response. If allies can't use the best US models, Chinese open-weight models may fill the void. The net result, experts argue, is less US technological influence and weaker allied defenses.

Going deeper

This compresses the bigger export-control dilemma: tighten too much and you lose allies and markets; loosen too much and frontier capabilities leak to adversaries. With security threats like the Agentjacking exploit (covered above) on the rise, the urgency of defensive capability adds weight to the letter. On the other side, hawks still warn that frontier models in adversary hands fuel offense too.

Bottom line

The "loosen-to-strengthen" letter shows that AI-era export controls can't be solved by tightening alone. The real question is where to draw the line between enabling defense and containing proliferation.