Letter: US cybersecurity leaders urge the White House to lift the ban on Anthropic's Mythos 5 and Fable 5, arguing the move hurts defenders more than attackers
The reported appeal follows U.S. restrictions on access to Anthropic’s Mythos 5 and Fable 5, including limits affecting foreign nationals and Anthropic staff in the U.S. Related coverage depicts the restrictions as an unusually direct form of government control over a model provider and as operationally disruptive after Anthropic disabled the models on short notice.
The subsequent coverage indicates that Anthropic and the U.S. were moving toward a safeguard under which the company would proactively detect and address security risks. That makes the cybersecurity leaders’ argument consequential: it frames access restrictions not only as an AI-safety measure, but as a constraint on defensive security work.
First-order effects
Cybersecurity defenders relying on Mythos 5 or Fable 5 face reduced access while the ban remains, whereas the White House faces organized pressure to distinguish defensive use from the security risks it says require restrictions.
Anthropic must operate under restrictions that affect its workforce and product availability, while demonstrating that its proposed safeguards can satisfy U.S. security concerns.
Second-order effects
A safeguard-based resolution would push Anthropic and other frontier-model providers toward more granular access controls, monitoring, and risk-response commitments rather than blanket availability or blanket exclusion.
Security teams may be forced to substitute other tools or workflows during the restriction, increasing the practical importance of whether competing models can support comparable defensive use under U.S. policy constraints.
Third-order effects
If security policy increasingly governs access to advanced AI by user category or perceived risk, model deployment may become a negotiated compliance function involving providers and government rather than a standard product-release decision.
The dispute exposes a durable tension in AI governance: controls intended to limit misuse can also limit defenders’ access, creating pressure for rules that can separate authorized defensive use from broader access without relying solely on broad bans.
The trend: This is one data point in the shift from general AI-safety debate toward operational U.S. controls over who can use frontier models and under what safeguards.
To support cyber defenders, ensure security of our critical infrastructure, and maintain America's AI leadership, it's time to #freefable https://freefable.org/
I wrote about what was actually in that #Fable guardrail bypass research paper, and why it should never have triggered an #AI model export control. We can't export control our way to cyber resilience. So many tshirt ideas. https://www.lutasecurity.com/ ... [image]
A lot of people are talking about the export control of Fable and why it was export-control-ed without actually mentioning the person who read the report, so this is straight from the source
David Sacks: “It's difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not “serious."" I think I fathom it now: [image]
Sounds like some folks at the WH were unaware Fable has >0 cyber abilities so thought something unsurprising was surprising, Dario rightly thought this was a misunderstanding, no one in govt who knew what was going on was looped in or could stop it, and ppl outside egged it on
Security people are signing an open letter asking the US government to remove the export restrictions on Fable/Mythos. I've read it multiple times and I still don't understand the argument. It seems like it's just a friends-of-Anthropic letter. [image]
“this action has taken the best models away from defenders, created market uncertainty, and risked America's AI leadership without any real risk to justify it.” “We, the undersigned executives and technical leaders from across the United States, write to you to ask you to lift [i…
Analysis of the jailbreak of Fable by someone who was asked to review the report. Indicates that the jailbreak allowed Fable to find deliberately planted vulnerabilities in test code that other models could find without jailbreak. Prompts were defensive in nature www.lutasecuri…