Sen. Richard Blumenthal, D-CT, said at a recent press conference that so-called “Big Tech” companies are “robbing America blind” by using creative works and user data to train generative artificial intelligence (AI) models like ChatGPT.
“Big Tech is robbing America blind with zero recourse to the artists, writers, creators, and consumers whose data and works are used to train AI models that then generate huge profits for Big Tech.” Blumenthal said at a press conference for a bill he is co-sponsoring with Sen. Josh Hawley, R-MO, that would create a federal tort allowing individuals to sue AI companies who use data without consent.
Under the “AI Accountability and Personal Data Protection Act,” which Blumenthal and Hawley recently introduced, AI systems, defined as “any data system, software, hardware, application, tool, or utility that operates, in whole or in part, using artificial intelligence” would have to obtain consent from users before using their data for training purposes.
Data covered by the bill includes any information that “is capable of being associated with, or can reasonably be linked, directly or indirectly, with a specific information” or is protected by copyright. It also includes personally identifying information, geolocation and biometric data, and online behavioral data such as search history or browsing patterns.
“Our bill would ban the use of copyrighted material or personal material without consent and if there is not consent, people would have the right to sue.” Blumenthal said. “They don’t right now. And if there is consent, there has to be notice of any sharing of that information with third parties.”
Blumenthal cited two recent rulings from California district judges that ruled partly against authors whose books were used to train AI models. In one case, a judge ruled that Anthropic’s purchase of books to train it’s AI model Claude were covered by fair use. The judge ruled that books that the company pirated in order to train Claude were not covered by fair use. Authors may receive damages for use of their pirated works.
In a second case, a judge also found that Meta’s use of books allegedly downloaded from “shadow libraries” was covered by fair use. Both cases found there was no direct empirical evidence showing authors’ markets had been harmed by being used to train AI models. Those cases are still ongoing.
Blumenthal said that the cases give Big Tech “total license to continue stealing America blind.”
The bill would require companies that intend to use user-generated data to train AI models to obtain express prior consent. If companies violate this, they could be forced to pay three times the amount of money earned from the use of personal data.
Blumenthal was joined by Vahid Behzadan, head of the Connecticut AI Alliance, who said the bill is “deeply pro-innovation and pro-business” and would fill gaps in existing legal protections, and Rock Vitale of AI implementation firm Easie,who said that Big Tech companies are currently “participating in a race to the bottom” with their AI model training methods.
Data privacy and AI regulation have been a priority for Democrats in the legislature in recent years, with Sen. James Maroney, D-Milford, leading a charge to have Connecticut become the first state in the nation to pass a framework for regulating AI.
For two consecutive years, the legislature has failed to advance SB 2, an omnibus AI bill championed chiefly by Maroney. A narrower governor’s bill brought forward by Gov. Ned Lamont, who has opposed Maroney’s bills, also failed to pass. Several AI-focused measures were included in the biennial budget, including a law that outlaws “synthetically created” revenge porn images. Another bill requires online services to notify users if their data will be used to train AI models.
Was this article helpful?
Republish our articles for free, online or in print, under a Creative Commons license.
