On Thursday, Anthropic revealed specialized AI models that were designed for American national security customers. The company has published “Claude GOV” models that were built in response to direct feedback from government customers to process operations such as strategic planning, intelligence analysis and operational support. The adapted models are reportedly already operating American national security authorities, with access limited to those who work in classified environments.
The Claude GOV models differ in different ways of Anthropic consumer and business offers, also known as Claude. They reportedly treat classified material, “less refusing” when entering into classified information and have been adjusted to handle intelligence and defense documents. The models also contain some anthropic “improved skill” in languages and dialects that are crucial for national security activities.
Anthropic says that the new models underwent the same “safety tests” as all Claude models. The company has pursued government contracts because it is looking for reliable sources of income in November, working together with Palantir and Amazon Web Services to sell AI tools to Defense customers.
Anthropic is not the first company that offers specialized chatbot services for intelligence services. In 2024, Microsoft launched an isolated version of OpenAI's GPT-4 for the American intelligence community after 18 months of work. That system, which worked on a special network of the government without internet access, became available for around 10,000 people in the intelligence community for testing and answering questions.
