Nearly all the massive AI information this 12 months was about how briskly the know-how is progressing, the harms it’s inflicting, and hypothesis about how quickly it’s going to develop previous the purpose the place people can management it. But 2024 additionally noticed governments make vital inroads into regulating algorithmic techniques. Here is a breakdown of crucial AI laws and regulatory efforts from the previous 12 months on the state, federal, and worldwide ranges.
State
U.S. state lawmakers took the lead on AI regulation in 2024, introducing tons of of payments—some had modest targets like creating research committees, whereas others would have imposed severe civil legal responsibility on AI builders within the occasion their creations trigger catastrophic hurt to society. The overwhelming majority of the payments didn’t cross, however a number of states enacted significant laws that might function fashions for different states or Congress (assuming Congress ever begins functioning once more).
As AI slop flooded social media forward of the election, politicians in each events acquired behind anti-deepfake legal guidelines. More than 20 states now have prohibitions towards misleading AI-generated political ads within the weeks instantly earlier than an election. Bills aimed toward curbing AI-generated pornography, significantly pictures of minors, additionally acquired robust bipartisan help in states together with Alabama, California, Indiana, North Carolina, and South Dakota.
Unsurprisingly, provided that it’s the yard of the tech business, a few of the most bold AI proposals got here out of California. One high-profile invoice would have compelled AI builders to take security precautions and held corporations chargeable for catastrophic damages brought on by their techniques. That invoice handed each our bodies of the legislature amid a fierce lobbying effort however was in the end vetoed by Governor Gavin Newsom.
Newsom did, nonetheless, signal greater than a dozen different payments aimed toward much less apocalyptic however extra fast AI harms. One new California legislation requires well being insurers to make sure that the the AI techniques they use to make protection determinations are honest and equitable. Another requires generative AI builders to create instruments that label content material as AI-generated. And a pair of payments prohibits the distribution of a lifeless particular person’s AI-generated likeness with out prior consent and mandates that agreements for dwelling peoples’ AI-generated likenesses should clearly specify how the content material can be used.
Colorado handed a first-of-its-kind within the U.S. legislation requiring corporations that develop and use AI techniques to take affordable steps to make sure the instruments aren’t discriminatory. Consumer advocates known as the laws an essential baseline. It’s probably that related payments can be hotly debated in different states in 2025.
And, in a center finger to each our future robotic overlords and the planet, Utah enacted a legislation that prohibits any governmental entity from granting authorized personhood to synthetic intelligence, inanimate objects, our bodies of water, atmospheric gases, climate, crops, and different non-human issues.
Federal
Congress talked lots about AI in 2024, and the House ended the 12 months by releasing a 273-page bipartisan report outlining guiding ideas and proposals for future regulation. But when it got here to truly passing laws, federal lawmakers did little or no.
Federal companies, alternatively, have been busy all 12 months attempting to satisfy the targets set out in President Joe Biden’s 2023 government order on AI. And a number of regulators, significantly the Federal Trade Commission and Department of Justice, cracked down on deceptive and dangerous AI techniques.
The work companies did to adjust to the AI government order wasn’t significantly horny or headline grabbing, however it laid essential foundations for the governance of private and non-private AI techniques sooner or later. For instance, federal companies launched into an AI-talent hiring spree and created requirements for accountable mannequin improvement and hurt mitigation.
And, in an enormous step towards growing the general public’s understanding of how the federal government makes use of AI, the Office of Management and Budget wrangled (most of) its fellow companies into disclosing vital details about the AI techniques they use that will impression individuals’s rights and security.
On the enforcement facet, the FTC’s Operation AI Comply focused corporations utilizing AI in misleading methods, equivalent to to write down faux opinions or present authorized recommendation, and it sanctioned AI-gun detection firm Evolv for making deceptive claims about what its product might do. The company additionally settled an investigation with facial recognition firm IntelliVision, which it accused of falsely saying its know-how was freed from racial and gender bias, and banned the pharmacy chain Rite Aid from utilizing facial recognition for 5 years after an investigation decided the corporate was utilizing the instruments to discriminate towards buyers.
The DOJ, in the meantime, joined state attorneys normal in a lawsuit accusing the actual property software program firm RealPage of a large algorithmic price-fixing scheme that raised rents throughout the nation. It additionally gained a number of anti-trust lawsuits towards Google, together with one involving the corporate’s monopoly over web searches that might considerably shift the steadiness of energy within the burgeoning AI search business.
Global
In August, the European Union’s AI Act went into impact. The legislation, which is already serving as a mannequin for different jurisdictions, requires AI techniques that carry out high-risk capabilities, equivalent to helping with hiring or medical selections, to bear danger mitigation and meet sure requirements round coaching information high quality and human oversight. It additionally bans using different AI techniques, equivalent to algorithms that may very well be used to assign a rustic’s residents social scores which are then used to disclaim rights and privileges.
In September, China issued a significant AI security governance framework. Like related frameworks printed by the U.S. National Institute of Standards and Technology, it’s non-binding however creates a typical set of requirements for AI builders to comply with when figuring out and mitigating dangers of their techniques.
One of probably the most fascinating items of AI coverage laws comes from Brazil. In late 2024, the nation’s senate handed a complete AI security invoice. It faces a difficult street ahead, but when handed, it might create an unprecedented set of protections for the sorts of copyrighted materials generally used to coach generative AI techniques. Developers must disclose which copyrighted materials was included of their coaching information, and creators would have the facility to ban using their work for coaching AI techniques or negotiate compensation agreements that will be primarily based, partially, on the scale of the AI developer and the way the fabric could be used.
Like the EU’s AI Act, the proposed Brazilian legislation would additionally require high-risk AI techniques to comply with sure security protocols.
