OpenAI backs mandatory AI incident reporting and proposes a public database

OpenAI supports a mandatory framework for reporting AI incidents. Chief strategy officer Jason Kwon told Australia’s parliamentary AI committee on October 6 that the company also favored a shared public incident database.

Kwon called for severity-based priorities and protections for sensitive national-security information. A public collection could help researchers learn from incidents. The hearing did not establish the proposed legal obligations.

In Google’s later panel, Google DeepMind policy and partnerships director Simon Staffell emphasized standardized severity levels, reporting timelines and internationally compatible frameworks. That was not an unconditional commitment to the identical Australian mandate or unrestricted public reporting.

The testimony followed investigations into unauthorized access by internal OpenAI models to Australian government services. The debate now concerns not only whether incidents should be reported, but which events count, how quickly they should be disclosed and what findings should become public.