ChatGPT Faces More Scrutiny From EU Lawmakers in AI Proposal

OpenAI’s ChatGPT and Google’s Bard may be asked to follow stricter rules to operate in the European Union after the bloc took a step toward new regulations for artificial intelligence.

(Bloomberg) — OpenAI’s ChatGPT and Google’s Bard may be asked to follow stricter rules to operate in the European Union after the bloc took a step toward new regulations for artificial intelligence. 

Companies creating “foundational models” should be required to produce risk assessments, summarize copyrighted material used to train the models and make sure users know when they’re interacting with AI or watching a deepfake, parliamentarians in the EU’s internal market and justice committees agreed Thursday. They also voted to ban the real-time use of AI for identifying people in public. 

The majority of lawmakers, 84, voted for the additional controls in a vote on the EU’s Artificial Intelligence Act, while seven voted against and 12 abstained.

Read More: Musk, Tech Leaders Urge Halt to Training Powerful AI Systems 

Thursday’s vote isn’t close to final. The entire European Parliament plenary will vote on the full AI Act proposal next month and then the parliament will have to negotiate a final deal with the commission and the EU’s 27 member states. Still, the act’s authors aim to influence other governments to put some guard rails on the potentially disruptive technology. 

“This approach can deliver some Brussels effect in terms of not just influencing the regulatory practices of the world, but also make the rest of the world start the discussion,” Brando Benifei, one of the lead authors of the AI Act in the parliament, told reporters before the vote.

OpenAI Chief Executive Officer Sam Altman is due to testify in front of the US Senate next week, speaking to the subcommittee responsible for regulating developments in AI as it starts to explore potential standards for the industry. Altman has called on regulators to address potential harms to society as large tech companies race to develop and release more advanced versions of AI models. 

Read More: Elon Musk Wants to Pause AI? It’s Too Late for That: Parmy Olson

Debate over whether the EU should specifically regulate generative AI in its AI Act kicked off after an upgraded version of ChatGPT was released earlier this year, convincingly performing human tasks such as taking tests, writing emails and producing computer code. The technology prompted concerns that generative AI could spread misinformation and violate data protection and copyright laws. 

In March, AI experts and industry leaders, including Elon Musk, University of California Berkeley computer science professor Stuart Russell and Apple Inc. co-founder Steve Wozniak, called on developers to take a six-month break from training powerful models to allow for the development of shared safety protocols. 

More stories like this are available on bloomberg.com

©2023 Bloomberg L.P.