Coinbase and Binance are seeking access to Anthropic’s Mythos artificial intelligence model as top crypto exchanges escalate efforts to counter emerging AI-driven security threats. The move underscores the dual-use nature of advanced AI: a powerful tool for defense that also poses significant risks if misapplied.
Exchanges eye advanced AI for security operations
As the crypto sector contends with increasingly sophisticated attacks, leading platforms are exploring cutting-edge models to strengthen detection and response. Access to Mythos could support:
- Real-time monitoring for on-chain anomalies and suspicious transaction patterns
- Enhanced phishing, fraud, and social engineering detection across user touchpoints
- Automated analysis and auditing of smart contracts and application code
- Threat intelligence enrichment and faster incident response workflows
Large language models and related AI systems are being integrated across security stacks to scale defenses, reduce false positives, and prioritize high-risk alerts. For global exchanges, even incremental gains in accuracy and speed can materially improve user protection and platform resilience.
Dual-use risks complicate adoption
The same capabilities that strengthen defenses can also be weaponized. Advanced models can lower the barrier to crafting convincing scams, generating malware, or identifying vulnerabilities at scale. This dual-use challenge is prompting exchanges to weigh tighter access controls, internal red-teaming, and robust governance over model use.
Balancing performance with safety remains central to any deployment. Firms are expected to emphasize restricted access, rigorous monitoring, and clear escalation procedures to mitigate misuse risks.
Industry outlook
The pursuit of Mythos access reflects a broader race within crypto to leverage AI for security, compliance, and operational efficiency. With billions of dollars lost to hacks and scams in recent years, the sector is prioritizing investments that can measurably reduce risk while maintaining strong user protections.
If secured, access to models like Mythos would likely be deployed behind strict controls and evaluated against tangible security outcomes, including reduced incident rates, faster detection times, and improved user safety metrics.