Anthropic today released the public version of its Mythos AI model, now called Claude Fable, according to a report on The Information. The model was first introduced as Claude Mythos Preview in April 2026 through Project Glasswing, with partners including AWS, Microsoft, Apple, and CrowdStrike. The public rollout comes with enhanced safeguards but retains the ability to autonomously discover and chain zero-day exploits across major operating systems and browsers — a capability that could reshape security for crypto firms.
From Mythos Preview to Fable
The preview version was already turning heads. It demonstrated the ability to find vulnerabilities in smart contracts, node software, and infrastructure, compressing what used to be weeks of manual auditing into hours. Polymarket assigned a high probability to a June 9 debut ahead of the official announcement, based on reports and leaks. Now that Claude Fable is live, the security community is watching how quickly it gets adopted.
Safeguards and pricing
Anthropic says the public release features enhanced safeguards to limit misuse while keeping strong reasoning, coding, and defensive security capabilities. The tradeoff comes at the price tag: access is expected to be roughly double current Claude Opus tiers. The original Mythos preview was priced five times higher than Opus, so the public version is still a premium product — but cheaper than the early-access tier.
The AI's ability to rapidly identify subtle vulnerabilities in smart contracts and infrastructure could give crypto firms that adopt it a significant edge. Faster audits mean faster patching, and Mythos-class tools promise to catch exploits before they're used in the wild. For teams working on DeFi protocols, exchanges, or cross-chain bridges, that's a direct competitive advantage. The timing matters — exploits in crypto have been ticking up, and automated defenses are suddenly a lot more plausible.
Broader rollout and what's next
Anthropic raised $65 billion recently and filed a confidential IPO, signaling it's betting big on being critical infrastructure. Broader availability of Claude Fable could roll out in the coming weeks, with possible expanded Project Glasswing participation that includes crypto-native organizations. That means a handful of crypto firms may get early access beyond the initial partners. For now, the public launch is the headline: a powerful, expensive AI that can hack your system before you do.
