In the letter, shared exclusively with FedScoop, Sen. Mark Warner, D-Va., urged the LOC’s copyright office to expand an existing “good-faith security research exemption” to include research that exists outside of traditional security concerns, such as bias, arguing that it would be the best path for ensuring a “robust security ecosystem” for tools such as generative AI. 

The letter from Warner, co-chair of the Senate Cybersecurity Caucus, is in response to a petition from Jonathan Weiss, founder of the IT consulting firm Chinnu Inc., that asked the LOC to establish a new exemption to address security research on generative AI models and systems. 

A spokesperson for Warner said in an email that an expansion to the exemption rather than an entirely new exemption “is the best way to extend the existing protections that have enabled a robust cybersecurity research ecosystem to the emerging issues surrounding safe AI.”

Warner’s letter mirrors a Department of Justice response to the same petition last month. The Computer Crime and Intellectual Property Section of the DOJ’s Criminal Division wrote that “good faith research on potentially harmful outputs of AI and similar algorithmic systems should be similarly exempted from the DMCA’s circumvention provisions.”

Said Warner: “It is crucial that we allow researchers to test systems in ways that demonstrate how malfunctions, misuse and misoperation may lead to an increased risk of physical or psychological harm.”

The Virginia Democrat, who has introduced bipartisan legislation on artificial intelligence security and emerging tech standards, pointed to the National Institute of Standards and Technology’s AI Risk Management Framework to acknowledge that AI’s risks “differ from traditional software risks in key ways,” opening the door for not only security vulnerabilities but also dangerous and biased outputs. 

The use of generative AI for fraud and non-consensual image generation are among the deceptive practices Warner noted as reasons for consumer protections, such as watermarks and content credentials. Additionally, the lawmaker asked the LOC to ensure that the potential expanded exemption “does not immunize” research that would intentionally undermine protective measures. 

“Absent very clear indicia of good faith, efforts that undermine provenance technology should not be entitled to the expanded exemption,” Warner said. 

The senator also asked the LOC to include security and safety vulnerabilities, especially involving bias and additional harmful outputs, in its expanded good-faith security research definition.

In response to Warner’s letter, Weiss said in an email to FedScoop that he doesn’t “care whether the existing exemption is expanded to include research on AI bias/harmful output, or whether an entirely new exemption is created. Our main concern is to secure protections for good faith research on these emerging intelligent systems, whose inner workings even the brightest minds in the world cannot currently explain.”

The Weiss petition and letters from DOJ and Warner were prompted by the LOC Copyright Office’s ninth triennial rulemaking proceeding, which accepts public input for new exemptions to the DMCA.

The post Senate Democrat pushes for expansion to copyright act to include generative AI research appeared first on FedScoop.

The system, which is called the Intelligent Records Consolidation Tool, was built with the help of an IT consulting group called the Savan Group and is maintained by the agency’s Justice Management Division. 

The software is supposed to assist in measuring the similarity of record schedules in order to reduce the time spent by the records manager, according to a description in the DOJ’s artificial intelligence inventory. It’s one of just four AI use cases mentioned in the inventory, which is required by a 2020 Trump administration executive order. 

While documents obtained by FedScoop show that the tool had an “informal kick-off” in late June 2020, the system is still in a pilot stage, a DOJ spokesperson confirmed. The AI inventory describes the tool as having been in production for “more than 1 year” and also states that the tool has been used in “multiple information management domains for the past three years.” 

Meanwhile, it’s not clear if there’s been any recent discussion about the tool within the agency.  A FedScoop public records request for documents related to the tool were all dated in 2020 and 

and a DOJ spokesman did not answer FedScoop questions about why the tool had not progressed from the pilot stage.

The slow expansion is notable, particularly as the Biden administration pushes federal agencies to adopt AI to improve internal operations. A DOJ artificial intelligence strategy, dated December 2020 and listed on the Justice Management Division’s page, also states that the agency hopes to “promote successful use cases, pilots, proof-of-concepts, and knowledge sharing to accelerate the deployment and appropriate use of AI.”

The Savan Group did not respond to a request for comment.

“The Intelligent Records Consolidation Tool as [sic] an administrative tool being used in a pilot stage to explore potential efficiencies and quality improvements for records consolidation and categorization,” Wyn Hornbuckle, a spokesperson for the agency, said in an email to FedScoop. “The use of technology such as artificial intelligence (AI) to increase automation in record processing is an emerging and promising area, which the department continues to explore while ensuring that there is sufficient human monitoring and appropriate safeguards are established.”

Madison Alder contributed to this article. 

The post DOJ AI tool has been in pilot stage for over three years appeared first on FedScoop.