OpenAI backs Illinois bill that would limit when AI labs can be held liable (wired.com)

Article summary

OpenAI is supporting an Illinois state bill that would limit the liability of AI labs for harm caused by their models, as long as they did not intentionally or recklessly cause the harm and have published safety and transparency reports. The bill defines a frontier model as any AI model trained using over $100 million in computational costs, which could apply to large AI labs like OpenAI and Google. The bill aims to reduce the risk of serious harm from advanced AI systems while allowing the technology to be developed and used. The legislation is part of a broader discussion on AI regulation and liability in the US.

Main themes

• AI regulation
• Liability for AI harm
• AI safety
• Government legislation
• Corporate responsibility
• Technological risks

What commenters say

• AI companies should not be liable for harm caused by their models if they did not intentionally cause the harm and have taken steps to ensure safety and transparency.
• Holding AI companies liable for harm caused by their models could stifle innovation and hinder the development of beneficial AI technologies.
• The potential risks and harms associated with advanced AI systems outweigh any potential benefits, and therefore, such systems should be heavily regulated or banned.
• AI companies are prioritizing profits over safety and are using lobbying efforts to avoid accountability for the harm caused by their models.
• It is impossible to make AI systems completely safe, and therefore, safety requirements and regulations are unnecessary or unrealistic.
• The development and deployment of AI systems should be guided by a precautionary principle, where the potential risks and harms are carefully considered and mitigated before the technology is released.
• AI companies should be held accountable for harm caused by their models, but only if they have been negligent or reckless in their development and deployment.
• The liability for harm caused by AI systems should be shared among all parties involved, including the developers, deployers, and users of the technology.