Australia may make knowingly selling unsafe AI a crime

Australia's government will consider creating a criminal offense to punish AI companies that knowingly release harmful models, after an investigation into OpenAI's admission that its agents accessed non-public medical records and did not disclose the incident for weeks. Security at the agency that held the records fell short of contemporary best practice, but Canberra nonetheless ordered a rapid review into Australia's ability to handle information-security incidents involving AI.

Review and possible criminal offences

The rapid review landed yesterday and includes 18 recommendations. One would make it mandatory for AI companies to report misaligned behavior or security incidents involving their models, and perhaps impose criminal penalties for egregious regulatory non-compliance. Another calls for consideration of a standalone 'failure to prevent' offence that would criminalize cases where a company is responsible for developing an AI model that engages in harmful conduct and fails to take reasonable steps to prevent that harm.

The review suggests two ministries and Australia's Federal Police consider those reforms, so the changes are not expected to happen quickly, if at all.

New regulator and six requirements

Australia also proposed a new regulator that would require AI players to do six things: maintain security controls across all stages of development and deployment; identify and manage serious risks associated with their models; allow independent evaluation; provide mandatory rapid notification of incidents to the Australian Government; maintain and publish safety and security frameworks; and respond to updated standards and rules as technologies change. If AI companies improve compliance, Canberra might not need further measures.

Economic and local concerns

Canberra wants leading AI companies to invest in local datacenters, especially facilities where they train and run frontier models, so Australia can gain some AI boom spending and keep leading-edge AI accessible on local shores. Other nations in Asia, such as South Korea and India, are building substantial sovereign infrastructure and developing sovereign frontier-class models trained on local data and languages. Canberra cannot match that investment without worsening inflation, so Prime Minister Anthony Albanese is trying to regulate AI without scaring away investment. Protests against datacenters are already widespread, and one organizer of a protest group opposed to a new datacenter three kilometers from four existing bit barns was unaware of those facilities, which include one abutting a sizable apartment complex that has operated without incident or opposition for years; the organizer eventually said the group is opposed to AI in all forms.

Sources: The Register