Humanity says it has released a new set of AI models tailored for US national security customers.
The new model, a custom set of the “Claude Gov” model, was “built to meet real operational needs based on direct feedback from government customers,” Humanity wrote in a blog post. Compared to Anthropic's consumer and business-centric model, the new custom Claude GoV model was designed to be applied to government operations such as strategic planning, operational support, and intelligence analysis.
“[These] The models are already deployed by agents at the highest level of US national security, and access to these models is limited to those operating in such a classified environment,” Humanity wrote in the post.[They] It went through the same rigorous safety tests as all Claude models. ”
Humanity is increasingly involved with U.S. government clients as they seek new, reliable sources of income. In November, the company collaborated with Amazon's cloud computing arm Palantir and AWS, a key human partner and investor, to sell human AI to defend its customers.
Anthropic says its new custom Claude GoV model better handles categorized material, better handles “rejecting” when involved in categorized information, and has a deeper understanding of the document within the context of intelligence and defense. The model also includes “improvement of proficiency” in languages and dialects, which are essential for national security activities, and “improvement of understanding and interpretation of complex cybersecurity data for intelligence analysis.”
Humanity is not the only top AI lab to chase defence contracts.
Openai is trying to establish close ties with the US Department of Defense, and Meta recently revealed that it has made its Llama model available to defense partners. Google is improving the version of Gemini AI that allows you to work within a categorized environment. Meanwhile, Cohere, which mainly builds AI products for businesses, is working with Palantir to develop AI models.