Why the US Patent-Trademark Office Has Banned Gen AI Use

Gen AI's propensity for 'bias, unpredictability and malicious behavior' has led the US Patent and Trademark Office to ban its use until such time as AI tools' performance has improved sufficiently to remove those risks. 

Given that the USPTO governs much of the legal protection for the country's innovation development, this is a strong warning about AI's current risks. JL

Reece Rogers reports in Wired:

The US Patent and Trademark Office banned the use of generative AI for any purpose last year, citing security concerns with the technology as well as the propensity of some tools to exhibit “bias, unpredictability, and malicious behavior.” USPTO staff are barred from relying on AI programs like OpenAI’s ChatGPT or Anthropic’s Claude for work tasks. The guidance also prohibits the use of any outputs from the tools, including images and videos generated by AI. Employees can use “state-of-the-art generative AI models” only inside the agency’s internal testing environment.

The US Patent and Trademark Office banned the use of generative artificial intelligence for any purpose last year, citing security concerns with the technology as well as the propensity of some tools to exhibit “bias, unpredictability, and malicious behavior,” according to an April 2023 internal guidance memo obtained by WIRED through a public records request. Jamie Holcombe, the chief information officer of the USPTO, wrote that the office is “committed to pursuing innovation within our agency” but are still “working to bring these capabilities to the office in a responsible way.”

Paul Fucito, press secretary for the USPTO, clarified to WIRED that employees can use “state-of-the-art generative AI models” at work—but only inside the agency’s internal testing environment. “Innovators from across the USPTO are now using the AI Lab to better understand generative AI's capabilities and limitations and to prototype AI-powered solutions to critical business needs,” Fucito wrote in an email.

 

Outside of the testing environment, USPTO staff are barred from relying on AI programs like OpenAI’s ChatGPT or Anthropic’s Claude for work tasks. The guidance memo from last year also prohibits the use of any outputs from the tools, including images and videos generated by AI. But Patent Office employees can use some approved AI programs, such as those within the agency’s own public database for looking up registered patents and patent applications. Earlier this year, the USPTO approved a $75 million contract with Accenture Federal Services to update its patent database with enhanced AI-powered search features.

 

The US Patent and Trademark Office, an agency within the Department of Commerce, is in charge of protecting inventors, awarding patents, and registering trademarks. It also “advises the president of the United States, the secretary of commerce, and US government agencies on intellectual property (IP) policy, protection, and enforcement,” according to the USPTO’s website.

At a Google-sponsored event in 2023, Holcombe, the author of the guidance memo, said government bureaucracy makes it difficult for the public sector to use new technologies. “Everything we do in the government is pretty stupid, when you compare it to the commercial world, right?” he said. Holcombe specifically cited cumbersome budgeting, procurement, and compliance processes, arguing that they hamper the government's ability to rapidly adopt innovations like artificial intelligence.

 

The USPTO is not the only government agency to ban staff from using generative AI, at least for some purposes. Earlier this year, the National Archives and Records Administration prohibited the use of ChatGPT on government-issued laptops, according to 404 Media. But soon afterward, the National Archives hosted an internal presentation that encouraged employees to “think of [Google’s] Gemini as a co-worker.” During the meeting, some archivists reportedly expressed concerns about the accuracy of generative AI. Next month, the National Archives is planning to release a new public chatbot for accessing archival records developed with technology from Google.

Other US government agencies are using—or avoiding—generative AI in different ways. The National Aeronautics and Space Administration, for example, specifically banned the use of AI chatbots for sensitive data. NASA did decide, however, to experiment with the technology for writing code and summarizing research. The agency also announced last week that it’s working with Microsoft on an AI chatbot that can aggregate satellite data to make it easily searchable. That tool is available only to NASA scientists and researchers, but the goal is to “democratize access to spaceborne data.”