The United Nations plans to form an advisory board that will explore the risks and opportunities presented by artificial intelligence.

The development comes as major tech industry players are also increasing their focus on AI governance. Against the backdrop of the U.N.’s Thursday announcement, OpenAI LP launched a new team tasked with identifying potential risks in its AI models. Earlier, the company and several other tech firms building cutting-edge neural networks made a set of commitments designed to ensure their software will be used safely.

The U.N.’s new AI advisory board, the High-Level Advisory Body on Artificial Intelligence, comprises 39 members. The participant list features experts from the government, civil society and academic sectors. Several prominent tech executives are also part of the board, including OpenAI Chief Technology Officer Mira Murati, Google LLC Senior Vice President James Manyika and Natasha Crampton, Microsoft Corp.’s chief responsible AI officer.

The board will focus on “building a global scientific consensus on risks and challenges, helping harness AI for the Sustainable Development Goals, and strengthening international cooperation on AI governance,” the U.N. said. The panel will also create recommendations for policymakers as part of its work.

As of Thursday, the board was scheduled to hold its first meeting today. Its 39 members will draft a set of preliminary recommendations on AI governance later this year and finalize the document by the summer of 2024. The recommendations will help shape the agenda of an upcoming U.N. conference, the Summit of the Future, where world leaders will explore the risks posed by AI along with other topics.

On Thursday, OpenAI likewise announced plans to make AI governance a bigger priority. In conjunction with the formation of the High-Level Advisory Body on Artificial Intelligence, the company announced a new team called Preparedness that will be tasked with exploring AI-related risks. It’s led by MIT computer science professor Aleksander Madry, who reportedly joined OpenAI in May.

The team’s responsibilities will include evaluating the capabilities of new OpenAI models and performing so-called red team exercises. Such exercises are designed to uncover ways that software systems, in this case neural networks, can be misused by hackers. 

The Preparedness team also plans to develop what OpenAI describes as a risk-informed development policy. This policy will outline how the company will go about evaluating the capabilities of its AI models, monitoring them for risks and addressing those risks. In addition, the document is expected to establish guidelines for “accountability and oversight” in OpenAI’s internal AI development workflows.

In conjunction with the formation of the new team, the company is launching an initiative it calls the AI Preparedness Challenge. It will allow members of the public to submit ideas about how hackers might go about misusing OpenAI’s models. According to OpenAI, the participants with the ten best submissions will receive $25,000 worth of credits that can be used to access its AI application programming interfaces.

Earlier, OpenAI and six other companies developing foundation models made a set of commitments to address the risks posed by their software. The commitments, which were made as part of a White House initiative announced in June, will see the companies test their AI models for safety issues. OpenAI and the other participants will also take steps to secure internal technical data, as well as publish reports about their models’ capabilities. 

Image: Unsplash