Cyberattacks, murder and loss of control: Inside Massachusetts' AI bill - Cape Cod Times

Cyberattacks, murder and loss of control: Inside Massachusetts' AI bill

- The Massachusetts state Senate is debating a bill to regulate artificial intelligence.

- This legislation is part of a larger $325 million economic development bill.

- The bill requires large AI developers to identify and mitigate potential catastrophic risks.

The Massachusetts state Senate debated a bill this week that it says would put guardrails on artificial intelligence.

The legislation is included in the Senate’s $325 million economic development bill, which provides money for certain industries like the defense sector and includes other legislation like putting age limits on e-bikes and requiring automatic two-family zoning.

The bill authorizes $75 million to “support the development and application of AI,” a press release said, while requiring safety frameworks for major AI developers.

Massachusetts lawmakers have been wrestling with the economic benefits of embracing artificial intelligence with the potential safety risks. Lawmakers have largely stayed away from sweeping regulations on artificial intelligence, and Gov. Maura T. Healey announced in February that the state’s workforce would start using a ChatGPT-powered assistant. But lawmakers have introduced many bills related to AI this session, including two House-passed bills regulating the use of AI in political advertisements.

What are the AI guardrails?

The bill requires large AI developers to identify any risk of potential catastrophe their AI model could cause, like the death of over 50 people or damages of more than $1 billion from doing things like engaging in a cyberattack or helping someone create a nuclear weapon. They must then develop a framework to mitigate the potential for these types of risks.

Under the bill, the attorney general would be allowed to sue an AI developer who violates the safety requirement.

The bill would also create a commission, led by the attorney general or a designee, to study the potential risks posed by artificial intelligence as well as the safety protocols designed by the developers.

What is 'catastrophic risk' in the bill language?

Catastrophic risk is defined as "foreseeable and material risk" of any development, storage or use of AI that could kill at least 50 people or cause $1 billion in damage in a single incident. The bill specifically says that risk could include giving expert-level assistance in creating or releasing "chemical, biological, radiological or nuclear weapon," cyberattacks, as well as murder, assault, extortion, or theft. It also covers whether the AI might "evade the control" of its developer or user.

See the latest list of adopted and rejected amendments, and read the full bill language and definitions, at https://malegislature.gov/Bills/194/S3178/BillHistory.