Anthropic’s AI Negotiations with US Defence Department

Anthropic is in the spotlight for its decision to decline an AI agreement with the US Defence Department, citing concerns over the reliability of AI for mass surveillance and autonomous weapon systems. Recently, CEO Dario Amodei has returned to discussions to potentially secure a deal with the US Defence Department, amidst concerns of being excluded from the military’s supply chain.

According to a report by the Financial Times, the company is in conversations with Emil Michael, the under-secretary of defence for research and engineering, regarding an agreement that would outline how the Pentagon is permitted to utilise Anthropic’s AI models.

If a deal is successfully concluded, the US military will continue access to Anthropic’s AI tools, thereby likely evading the detrimental “supply chain risk” label that has loomed over the company. Last week, Defence Secretary Pete Hegseth hinted at the possibility of applying this label to Anthropic, although it has not yet been officially designated.

Safeguards Maintained by Anthropic

Anthropic established clear boundaries last week concerning the deployment of its AI technologies by the US military. The organisation stood firm against requests to eliminate two key safeguards from their contract with the Pentagon: a prohibition on the use of its models for extensive domestic surveillance and a limitation on the use of fully autonomous weapon systems.

Dario Amodei’s Insights on AI Reliability

In a blog entry, Amodei emphasised that AI should not be relied upon without human oversight. Currently, the company is back in negotiations with the Pentagon, but no formal agreement has been reached as both parties await confirmation.

Despite potential challenges from its refusal to cooperate with the US military, the popularity of Anthropic’s Claude AI chatbot continues to rise. Conversely, OpenAI experienced significant backlash after establishing a deal with the Pentagon, leading to a social media campaign labelled “Cancel ChatGPT”, which resulted in a 295% increase in uninstallation last week.