Anthropic announced Claude Gov on Thursday, his product designed especially for the US Defense and Intelligence Agency. AI models have looser handrails for exploit by the government and are trained for better analysis of classified information.
The company stated that models that announce that “are already implemented by agencies at the highest level of US national security”, and access to these models will be restricted to government agencies that support classified information. The company did not confirm how long they were in exploit.
Claude GOV models are specially designed to uniquely satisfy government needs, such as threat assessment and intelligence analysis, in accordance with Anthropic Blog post. And although the company said that “has undergone the same strict security tests as all our Claude models”, the models have some specifications of national security work. For example, “they refuse less when they are involved in classified information,” which is provided in them, something that Claude is trained by consumers to mark and avoid.
Claude GOV models also have a better understanding of documents and context in defense and intelligence, in accordance with anthropic and better proficiency in languages and dialects related to national security.
The exploit of AI by government agencies has long been examined due to its potential damage and wave effects for minorities and sensitive communities. There was a long list of unlawful arrests Many US states Due to the exploit of facial recognition by the police, Documented evidence prejudices in predictive police and discrimination in government algorithms Evaluate assistance in the field of social welfare. Over the years, there is also controversy in the industry compared to gigantic technology companies, such as Microsoft, Google and Amazon, allowing military-part-time in Israel-using AI products, with campaigns and public protests as part of protests as part of public protests under No technology for apartheid movement.
Anthropics Use policy In particular, it decides that every user must “do not” create or facilitate the exchange of illegal or highly adjustable weapons or goods “, including the use of products or services of anthropics for” production, modification, design, sale or distribution of weapons, explosives, hazardous materials or other systems designed to cause damage or loss of human life. ”
At least eleven months ago the company he said He created a set contractual exceptions According to the exploit policy, which is “carefully calibrated to enable beneficial exploit by carefully selected government agencies.” Some restrictions – such as disinformation campaigns, design or use of weapons, construction of censorship systems and malicious cyberspace – will remain prohibited. But anthropic may decide to “adapt the restrictions on the mission and legal bodies of the government entity”, although it will “balance the beneficial exploit of our products and services with limiting potential damage.”
Claude Gov is Anthropica’s response to ChatgPT GOV, the OpenAI product for US government agencies, which he introduced in January. It is also part of the wider trend of AI giants and startups that want to strengthen their companies with government agencies, especially in an uncertain regulatory landscape.
When Opeli announced Gov Chatgpt, the company said that over the past year over 90,000 employees of federal, state and local governments used their technology to translate documents, generate abstracts, design of policy notes, code, building and others. Anthropic refused to divide the numbers or cases of using the same type, but the company is part of the Fedstart Palantir program, SAAS offers for companies that want to implement federal government software.
Scale AI, Gigant AI, which provides training data for industry leaders such as OpenAI, Google, Microsoft and Meta, signed a contract with the Defense Department in March for the first of its kind AI program for US military planning. And since then he has expanded his activity to global governments, recently concluding a five -year contract with a runny nose to provide automation tools for civil service, healthcare, transport and others.