The Australian government is preparing legislation that would make AI companies criminally liable for incidents of harm, a move prompted by a recent high‑profile breach involving OpenAI.

The Australian government is drafting legislation that would make artificial intelligence firms criminally liable if their systems cause harm, a response sparked by a recent breach involving OpenAI’s technology.

Background to the proposed law

The draft bill, known as the AI Safety and Accountability Act, seeks to impose criminal penalties on companies whose AI products result in physical injury, death, or significant property damage. Lawmakers argue that existing civil remedies are insufficient to deter negligent deployment of powerful AI systems.

Key provisions under consideration

  • Mandatory risk assessments before AI products are released to the public
  • Obligation to implement real‑time monitoring and rapid response mechanisms
  • Criminal fines up to $10 million or five years imprisonment for executives responsible for preventable harms

The legislation also proposes a new regulatory body to oversee compliance, granting it powers to order the suspension of AI services and to demand detailed technical documentation from developers.

Industry reaction

Technology firms, including OpenAI, have expressed concerns that criminal liability could stifle innovation and lead to overly cautious AI deployment. Critics warn that the law may be difficult to enforce given the complex, often opaque nature of AI decision‑making.

Consumer advocacy groups, however, welcome the move, saying it signals a commitment to protect citizens from unintended consequences of rapidly advancing AI technologies.

For ongoing coverage of AI policy developments, see Technology section on uxc.news and the latest updates on OpenAI on Momentum.

"Holding AI firms criminally accountable could create a necessary safety net, but the law must be clear and proportionate," said a legal analyst specializing in technology regulation.

The draft bill is expected to be debated in Parliament later this year, with a public consultation period slated for early 2027.

Read the full report in Technobezz coverage of Australia’s AI criminal liability proposal.