DeFi Users Urged to Revoke Approvals Ahead of Anthropic Mythos AI Release

Rising AI Capabilities Pose New Risks to DeFi

Anthropic is set to broaden availability of its Mythos AI model, a system noted for its advanced ability to identify software vulnerabilities. This development has prompted urgent calls for decentralized finance users to review and revoke unnecessary token approvals.

Token approvals allow smart contracts to interact with user funds. Over time, these permissions can accumulate without notice, creating ongoing exposure if any approved contract faces compromise. Security analysts emphasize that models like Mythos could accelerate the discovery of weaknesses in DeFi protocols.

The DeFi Investor's Warning

Crypto analyst The DeFi Investor highlighted the capabilities of Mythos in a recent statement, noting its superiority in detecting severe vulnerabilities compared to previous models. The analyst referenced recent AI-assisted discovery of a critical bug in Zcash's shielded pool, which led to significant market impact.

This concern aligns with broader discussions in the crypto community about AI's impact on security. OpenZeppelin co-founder Manuel Aráoz previously stated that AI coding agents have made much of DeFi unsafe, advising caution with major protocols.

Anthropic's Mythos and Project Glasswing

Mythos has been available in restricted form since April through Project Glasswing, initially to around 50 organizations including major tech companies. The company plans to expand access to additional cybersecurity and research entities while maintaining safeguards.

Anthropic acknowledges the dual nature of these capabilities: they can strengthen defenses but also potentially empower attackers during a transitional period. A public version is expected with substantial guardrails.

AI-Powered Adaptive Threats Emerge

Recent research demonstrates AI worms capable of adapting attacks in real time. These systems use large language models to analyze targets, generate custom exploits, and propagate across networks without human intervention. Running locally on infected machines, they prove difficult to detect and stop using traditional methods.

In controlled tests across mixed operating systems with common vulnerabilities, the AI worm adjusted tactics effectively. This marks a shift toward autonomous generative adversaries, raising concerns for blockchain infrastructure, exchanges, and self-custody solutions.

Recent Incidents Highlight Vulnerabilities

Several high-profile incidents underscore the current risks:

While some experts like Aave Chan Initiative founder Mark Zeller argue that AI-related fears may be overstated, with most past failures stemming from other factors, the evolving threat landscape requires heightened vigilance.

Market Context and Altcoin Season Outlook

The security discussion occurs against a backdrop of cautious market sentiment. The Altcoin Season Index currently stands at 49, indicating Bitcoin dominance rather than broad rotation into alternative assets. Analysts point to tight liquidity conditions and historical patterns showing altcoin rallies are infrequent and short-lived.

With thousands of tokens competing for capital, significant inflows into altcoins may require improved market conditions and stronger monetary easing signals.

Best Practices for DeFi Participants

Beyond immediate approval revocation, users should prioritize:

Anthropic positions its work as contributing to long-term defensive advantages, but acknowledges the near-term challenges. As AI capabilities advance, the crypto ecosystem must adapt its security practices accordingly.

The coming months represent a critical stress test for DeFi. Users who proactively manage risks will be better positioned as the AI technology evolves.