The Pentagon and AI: Questioning Surveillance of Americans

A Complex Debate on AI Surveillance

The current debate between the U.S. Department of Defense and the artificial intelligence company Anthropic highlights a complex legal question: does the U.S. government have the right to surveil its citizens using artificial intelligence? This inquiry, while crucial, does not have a simple answer. More than a decade after Edward Snowden's revelations about the NSA's mass collection of phone metadata, the United States continues to navigate a gray area between public perceptions and legal authorizations.

The Conflict Between Anthropic and the Pentagon

The point of contention between Anthropic and the government lies in the potential use of Anthropic's AI Claude by the Pentagon to analyze large amounts of commercial data on American citizens. Anthropic has firmly stated that its AI should not be used for mass domestic surveillance or for the development of autonomous weapons, which are capable of targeting and killing without human intervention. Following the breakdown of negotiations, the Pentagon classified Anthropic as a supply chain risk, a designation typically reserved for foreign companies perceived as threats to national security.

OpenAI and Its Initial Agreement with the Pentagon

On its part, OpenAI, the company behind the famous ChatGPT, had initially entered into an agreement with the Pentagon allowing the use of its AI for "all legal uses." This wording drew criticism, with some arguing that it left the door open for domestic surveillance. In response, a massive uninstall movement of ChatGPT took place, accompanied by protests outside OpenAI's headquarters in San Francisco, where demonstrators questioned the company's boundaries.

OpenAI's Revision of the Agreement

In the face of public pressure, OpenAI announced a revision of its agreement, ensuring that its AI would not be used for domestic surveillance. The company also clarified that its services would not be made available to intelligence agencies such as the NSA. OpenAI's CEO, Sam Altman, stated that current legislation prohibits domestic surveillance by the Department of Defense, and that their contract simply reflected this legislation.

Diverging Opinions on the Legality of Surveillance

Dario Amodei, CEO of Anthropic, expressed a differing opinion, asserting that if such surveillance is currently legal, it is only because legislation has not yet caught up with the rapid advancements in AI. This divergence raises the question of whether the current legal framework is suitable for modern technological capabilities.

The Legal Definition of Surveillance

Whether the Pentagon can legally surveil Americans using AI depends on the very definition of surveillance. According to Alan Rozenshtein, a law professor at the University of Minnesota, many practices that the general public considers surveillance are not necessarily recognized as such by law. This includes access to public information like social media posts, surveillance camera footage, and voter registration records, as well as data obtained incidentally while surveilling foreign nationals.

The Government's Purchase of Commercial Data

The government also has the option to purchase commercial data from companies, which can include sensitive personal information such as mobile location and browsing histories. Agencies like ICE, the IRS, the FBI, and the NSA have increasingly turned to this data market, fueled by a digital economy that collects user data for advertising purposes. These datasets allow the government to access information that would normally only be available with a warrant or subpoena.

Legal Limits on Data Collection

Rozenshtein emphasizes that there is a considerable amount of information that the government can collect on American citizens without being constrained by the Constitution, particularly the Fourth Amendment, or by legislation. There are no significant restrictions on how the government can use this data.

The Evolution of Laws in the Face of New Technologies

Historically, the Fourth Amendment was designed at a time when information collection involved physical intrusion into homes. Subsequent laws, such as the Foreign Intelligence Surveillance Act of 1978 and the Electronic Communications Privacy Act of 1986, were enacted at a time when surveillance primarily involved listening to phone calls and intercepting emails. Most laws governing surveillance were established before the rise of the Internet, when people did not generate vast amounts of online data and the government did not have sophisticated tools to analyze this data.

The Impact of AI on Surveillance

Today, artificial intelligence significantly amplifies surveillance capabilities. It allows for the aggregation of individual information fragments to identify patterns, draw inferences, and construct detailed profiles of individuals on a large scale. As long as the government collects this information legally, it can use it as it sees fit, including integrating it into AI systems. Rozenshtein concludes that legislation has not yet caught up with these technological advancements.