World needs to ‘get ready’ for more powerful AI, Anthropic co-founder says
Speaking on Monday at the Semafor World Economy, Jack Clark, Anthropic co-founder and head of Public Benefit, discussed Project Glasswing’s mission and how its central model, Mythos, was developed.
Clark said the work to develop Mythos began with the realization that existing AI models have the potential to be trained to focus on cybersecurity operations and were headed toward cybersecurity-specific development.
“What we saw made us realize: the next time we train a really big model, we should expect it to have these capabilities sort of inherent to it, rather than ones that we need to elicit,” Clark said.
Even in its early stages of development, Clark said Mythos surpassed every benchmark that researchers applied and could detect vulnerabilities in applications and operating systems.
Despite this early capability coming from Anthropic, Clark said more Mythos-like models will emerge from other developers.
“This is not a special model,” he said. “There will be other systems just like this in a few months from other companies, and then a year to a year and a half later, there'll be open-weight models from China that have these capabilities. So the world is going to have to get ready for more powerful systems that are going to exist within it.”
Project Glasswing, which grants certain companies selective access to Mythos for further safety and capability testing, will help reveal more about the AI’s capabilities, Clark added.
Clark’s comments come as the Pentagon and Anthropic remain locked in a legal dispute over the Defense Department labeling the company a supply chain risk following its decision to prevent the Pentagon from using its models for domestic surveillance and lethal autonomous weapons.
]]>