Adam Satariano of the New York Times reports this evening that European Union policymakers have agreed on a new law, the A.I. Act, to regulate artificial intelligence, making it one of the world’s first comprehensive laws on AI.
While the law is subject to final approval and will not take effect for 12 to 24 months after,
- The Act aims to harness the benefits of AI while mitigating risks such as automation of jobs, misinformation spreading, and national security threats.
- The law targets AI’s riskiest uses by companies and governments, including law enforcement and operation of crucial services – water, gas and electric etc.
- Makers of large general-purpose AI systems, like the ChatGPT chatbot, will face new transparency requirements.
- Use of facial recognition software by police and governments will be restricted, with exceptions for safety and national security.
- Companies violating the regulations could face fines up to 7% of global sales.
- The Act will affect major AI developers like Google, Meta, Microsoft, and OpenAI, as well as businesses using AI in areas like education, healthcare, and banking. Governments using AI in criminal justice and public benefits allocation will also be impacted.
Though the law is being hailed as a regulatory breakthrough, its effectiveness is still up in the air as many policy aspects won’t take effect, like I said, for 12 to 24 months, a lifetime in AI development.
Enforcement of the Act also remains unclear, as it involves regulators across 27 nations and requires hiring experts at a time of tight government budgets. Legal challenges are obviously expected as well.