Pentagon May Blacklist Anthropic Over ‘Woke AI’ Bias Concerns
The US defence secretary, Pete Hegseth, is threatening to blacklist Anthropic. The reason is that the company has lowered its safety standards in order to continue doing business with the U.S. military.
The menace occurred as Hegseth and Anthropic CEO Dario Amodei were holding a meeting on Tuesday, as captured by two individuals who had firsthand knowledge of this meeting but were not designated to address the media.
Although the two parties accepted that Hegseth had promised to punish Anthropic in case it failed to yield to the demands of the administration, there are differences on what the threat entailed. One individual involved in the conversation said that Hegseth threatened to cancel the contract of Anthropic valued at 200 million dollars with the defence department, and another Pentagon employee said that repercussions would include compelling Anthropic to make its AI tools available to the federal government and blacklisting the company as a future employer of the U.S. military.
Hegseth has stated that Anthropic must permit the U.S. to employ its AI in all ways it considers legal, which may be the use of AI-controlled warfare and surveillance.
The Pentagon officials reported that the defence department is already intending to continue using the tools of Anthropic, though the company may not want this.
As the officials explain, in case Anthropic refuses to give way, Hegseth will invoke the Defence Production Act, which is a 1950s law that is typically invoked during times of national emergencies as a way of compelling companies to manufacture specific products deemed to be of national interest.
This would make Anthropic grant its equipment to the military to use or not to use depending on the wishes of a high-ranking official of the Pentagon.
The government action against Anthropic would also involve declaring the firm a supply chain risk, which is practically equivalent to blacklisting the firm by the government, as the individuals who attended the meeting put it.
The stance on domestic surveillance and AI weapons by Anthropic has been dubbed by Hegseth and other Trump officials as being ‘woke AI’. In 2019, White House AI czar David Sacks was one who drafted an executive order targeting tech companies on the claim.
According to AI experts, the term woke AI is an imprecise and vaguely defined one that Trump officials appear to employ to denote the existence of any and all safety controls on potent AI tools and the notion that AI chatbots are liberal biased and coded to think so.
The rival AI companies like OpenAI and Google have promised to enable their AI systems to be deployed in all situations that are legal, as has the xAI of Elon Musk, which this week was permitted to be used in classified environments.
Last summer, the Pentagon granted up to $200 million contracts to the four companies Anthropic, Google, OpenAI and xAI, although the first that passed through a security check to be allowed into the classified domain was Anthropic, which was deemed by the defence authorities as the most advanced and secure model to use when it came to sensitive military applications.
The controversy with the White House is particularly timed when Anthropic is in an increased level of scrutiny, given the fact that the company is to be pushed into a state of going public this year. The investors will be unsure of how well the friction with the administration will go.
