Few crypto firms gain access to frontier AI models as developers weigh security risks against escalating cyberattack capabilities.
Anthropic is restricting access to its latest AI models, citing concerns that attackers may exploit them faster than defenders can adapt. The controlled rollout aims to reduce potential damage during early testing phases, according to discussions at a recent industry forum.
The number of critical-severity vulnerabilities has increased following the launch of Claude Mythos Preview, data from Epoch AI shows. Industry leaders argue that while safeguards are necessary, legitimate defenders need faster access to advanced models to counter evolving threats.
Solana Foundation’s chief information security officer and Blockchain Capital executives support eventual broader access but emphasize the need for streamlined verification processes to ensure defenders can deploy AI tools effectively.