The US Commerce department are lifting the previous restrictions on Anthropic’s Mythos AI model, that were designed to limit access of this powerful system to foreign nations. Mythos has proven itself to be the next generation of models that are spectacular at identifying vulnerabilities and developing exploits to take advantage of them.

The problem is that OpenAI and Chinese firms are very close behind, making this a race of weeks before such technology is widely available.

The tech can be used for attacks if available, but must also be used by developers to fix weaknesses before being attacked.  It is a tough dilemma.

The entire regulatory model may need to be transformed to something like an N-1 designation, where the latest models are restricted, but one generation older is open for use.  This might be the balance needed to allow companies access to sufficient technologies for vulnerability management, while making it more difficult for foreign state actors to use the latest tech to victimize adversaries in ways that push their foreign policy, undermine economies, or put critical infrastructures at grave risks.

Striking such a balance must be designed to keep pace with rapid technology innovation without requiring slow review and oversight from legislatures or agencies. 

It is time to establish AI governance frameworks that continue to enable the advantages of innovation investment and successes, while still supporting capabilities for defensive structures to remain reasonably effective.