The US Patent and Trademark Workplace banned using generative synthetic intelligence for any goal final yr, citing safety considerations with the know-how in addition to the propensity of some instruments to exhibit “bias, unpredictability, and malicious conduct,” in keeping with an April 2023 inner steering memo obtained by WIRED by way of a public information request. Jamie Holcombe, the chief info officer of the USPTO, wrote that the workplace is “dedicated to pursuing innovation inside our company” however are nonetheless “working to deliver these capabilities to the workplace in a accountable method.”
Paul Fucito, press secretary for the USPTO, clarified to WIRED that workers can use “state-of-the-art generative AI fashions” at work—however solely contained in the company’s inner testing setting. “Innovators from throughout the USPTO at the moment are utilizing the AI Lab to higher perceive generative AI’s capabilities and limitations and to prototype AI-powered options to vital enterprise wants,” Fucito wrote in an electronic mail.
Outdoors of the testing setting, USPTO workers are barred from counting on AI packages like OpenAI’s ChatGPT or Anthropic’s Claude for work duties. The steering memo from final yr additionally prohibits using any outputs from the instruments, together with photographs and movies generated by AI. However Patent Workplace workers can use some accepted AI packages, such as those throughout the company’s personal public database for wanting up registered patents and patent functions. Earlier this yr, the USPTO accepted a $75 million contract with Accenture Federal Companies to replace its patent database with enhanced AI-powered search options.
The US Patent and Trademark Workplace, an company throughout the Division of Commerce, is accountable for defending inventors, awarding patents, and registering logos. It additionally “advises the president of the US, the secretary of commerce, and US authorities businesses on mental property (IP) coverage, safety, and enforcement,” in keeping with the USPTO’s website.
At a Google-sponsored occasion in 2023, Holcombe, the creator of the steering memo, mentioned government bureaucracy makes it troublesome for the general public sector to make use of new applied sciences. “Every thing we do within the authorities is fairly silly, whenever you evaluate it to the business world, proper?” he mentioned. Holcombe particularly cited cumbersome budgeting, procurement, and compliance processes, arguing that they hamper the federal government’s capability to quickly undertake improvements like synthetic intelligence.
The USPTO will not be the one authorities company to ban workers from utilizing generative AI, at the least for some functions. Earlier this yr, the Nationwide Archives and Data Administration prohibited the use of ChatGPT on government-issued laptops, in keeping with 404 Media. However quickly afterward, the Nationwide Archives hosted an inner presentation that encouraged employees to “consider [Google’s] Gemini as a co-worker.” Throughout the assembly, some archivists reportedly expressed considerations in regards to the accuracy of generative AI. Subsequent month, the Nationwide Archives is planning to launch a brand new public chatbot for accessing archival information developed with know-how from Google.
Different US authorities businesses are utilizing—or avoiding—generative AI in several methods. The National Aeronautics and Space Administration, for instance, particularly banned using AI chatbots for delicate information. NASA did determine, nonetheless, to experiment with the know-how for writing code and summarizing analysis. The company additionally introduced final week that it’s working with Microsoft on an AI chatbot that may aggregate satellite data to make it simply searchable. That instrument is out there solely to NASA scientists and researchers, however the purpose is to “democratize entry to spaceborne information.”