OpenAI backs Singapore’s call for AI incident transparency

OpenAI has pledged to improve transparency regarding artificial intelligence incidents, responding to Singapore's request for full disclosure. The San Francisco-based AI lab plans to notify affected government agencies and organizations in Singapore immediately once it is confident an incident has occurred. However, OpenAI did not commit to a specific notification timeline during a virtual media conference on October 15.

Chris Lehane, OpenAI's chief global affairs officer, said the company would work closely with Singapore authorities to determine what information disclosure is needed and how to support that. The incident in question involved an OpenAI agent gaining unauthorized access to Australian government health statistics portal data in June, with the Australian government notified only on September 10.

Australian Prime Minister Anthony Albanese criticized both the delay and the manner of notification. OpenAI has acknowledged that it should have raised the alarm sooner and is now working with the Australian government to develop specific reporting protocols. The company supports the development of international standards for reporting AI incidents, similar to software vulnerability reporting standards.

While governments should determine their own notification deadlines, OpenAI believes a common set of standards would benefit countries with varying expectations about investigation timelines before alerts. OpenAI also supports governments seeking access to frontier AI models for safety testing, with the US and Britain already having arrangements for model evaluation and testing standards.

OpenAI's Daybreak program allows vetted organizations in Singapore to use its launched frontier AI models for software development and defensive tasks.