As expected, President Donald Trump repealed his predecessor Joe Biden’s 2023 artificial intelligence (AI) regulations on his first day in office.
The former executive order would have required the federal government to vet the advanced AI models of major developers such as OpenAI, Google, Amazon and other tech giants. Biden’s order also established chief AI officers in major federal agencies and set out frameworks that addressed ethical and security risks.
Trump’s reversal marks a significant policy shift that is lighter on regulations and guardrails and more pro-growth and pro-innovation. However, it is unclear how his repeal of Biden’s executive order will be enacted on the ground for federal agencies that already have instituted such policies. Notably, Trump established the first executive order on AI during his first term in office.
As the U.S. rolls back AI guardrails and oversight, France critiqued the EU AI Act, the most comprehensive set of AI regulations globally. Any company doing business in the EU or interacting with EU citizens must comply with these regulations. The act focuses on gauging AI models and tools based on the different risk levels they pose to their users.
Clara Chappaz, France’s first minister of AI, said that rather than being a hurdle for companies, the act should be seen as a ‘tool to help you go faster.” Europe can unite to establish a framework that doesn’t hinder innovation by giving startups what they need in terms of infrastructure, access to data, and other support, she said at a panel during the World Economic Forum in Davos on Wednesday (Jan. 22).
Chappaz urged EU officials not to get “stuck” on the idea of regulating just to regulate, but rather see regulation as a springboard to innovation.
IBM CEO and Chairman Arvind Krishna, who was on the same panel, added that regulations should be “light touch.” But he pointed to the reality that even if the legislation’s intent was for regulations to be a “scalpel when it comes to the hands of regulations, it’s a sledgehammer.” He said there is a place for a heavy-handed approach: AI systems that pose extreme risk.
Krishna also believes that developers of AI models, mainly OpenAI and the Big Tech companies, are held accountable for what they are producing.
Mistral AI CEO and Co-Founder Arthur Mensch, a French AI startup that is Europe’s answer to OpenAI, took issue with tech executives cozying up to the U.S. government. During the same panel discussion at the World Economic Forum, Mensch said there should be a gap between the private and public sectors as legislators have their role to play.
Among notable tech executives who attended the Trump inauguration were Tesla CEO Elon Musk, Amazon Founder Jeff Bezos, Apple CEO Tim Cook, Google CEO Sundar Pichai, Meta CEO Mark Zuckerberg and Google Co-Founder Sergey Brin.
Mensch also disputed the idea that AI is costly to develop, a narrative that has served OpenAI and other AI startups as they raise billions of dollars for AI foundation model training. He said Mistral decided to develop open-source models to make AI widely available to everyone, noting that thus far, no fraudsters have used the models for malicious purposes. The biggest risk is the concentration of AI into a few hands, Mensch added.
The U.S. Department of Commerce has published guidelines to optimize the use of open data for generative AI models. The guidelines affect not only Commerce Department employees but also the general public. The department holds some of the nation’s most valuable open data assets, such as about the U.S. economy, people and the environment.
The Commerce Department said that it has updated the way it processes and publishes data to reflect technological innovations, from digitizing records to using machine-readable formats in the 2010s. Now, the department is preparing its data for use by generative AI models.
The new guidelines make its data not only machine-readable but also machine-understandable. “This means preserving the meaning and context of the data in ways that generative AI systems can accurately interpret and utilize,” the department said in its report.
Enabling generative AI models to better understand Commerce Department data will enable non-technical users to more easily get insights from the data by using AI chatbots like ChatGPT.
The document details five major focus areas: Documentation, data and metadata formats, data storage and dissemination, data licensing and usage, and data integrity and quality.