In the first of a three-part series on Anthropic’s new powerful Mythos AI model, we look at its impact on Chinese AI, cybersecurity and competition with the US.

US start-up Anthropic announced its latest artificial intelligence model Claude Mythos Preview on April 7, sparking an unprecedented global response among policymakers and regulators due to its powerful ability to identify and exploit cybersecurity vulnerabilities.

Instead of a public release, Anthropic released Mythos to a consortium of US companies including Cisco, JPMorgan Chase and Nvidia to use the model to secure their critical software in an initiative called Project Glasswing.

The US National Security Agency and other US governmental bodies have already begun using the model, according to US media reports. No Chinese companies were invited to use the model to patch their cybersecurity vulnerabilities given Anthropic’s labelling of China as an “adversarial nation”, while the company’s services continue to be banned in greater China, including Hong Kong.

How do Chinese models compare?

According to Beijing-based consultancy Concordia AI, China’s primarily open-source models still lag closed-source US models in their cyber capabilities. However, the rapid progress of Chinese models over the past year has meant that their capabilities are also improving rapidly.