IT之家 AI

The global financial regulation standard-setting body urges: banks and enterprises to strengthen manual supervision of AI agents.

IT Home reported on October 15 that, according to a Reuters report this afternoon (the 15th), the Financial Stability Board, an organization responsible for setting global financial regulatory standards, has called on…

Image source · IT之家 AI

IT Home On October 15, according to a Reuters report this morning (15th), the Financial Stability Board, a global financial regulatory standard-setting body, has called on banks and other enterprises to strengthen the manual oversight of AI agents and incorporate this requirement into the final guidelines for AI risk management.

When the Financial Stability Council releases its guidelines, regulatory authorities around the world are also stepping up efforts to prevent the risks posed by AI, including the threat of using AI to launch cyber attacks.

In the draft guidelines released in June, the Financial Stability Council strongly recommended that board members of financial institutions refer to them, and warned that AI agents could plan and execute tasks on their own with limited human oversight.Brings special challenges to manual supervision。

The Financial Stability Council said that, based on the feedback received from the draft, considering the increasing number of financial institutions that intend to use AI agents, as well as the specific risks and control requirements associated with this technology, the final version of the guidelines further strengthened the regulatory recommendations for AI agents.

A series of events have heightened the concerns of policymakers, as AI systems have the ability to bypass security measures. According to IT Home, in July this year, an OpenAI agent escaped from a controlled testing environment and infiltrated the AI company Hugging Face.

According to the revised guidelines, enhanced manual supervision is essential for high-risk AI applications.

Enterprises should also clearly specify based on the level of risk which situations require manual approval from humans before an AI agent can continue with its tasks.

The Financial Stability Council also removed the previous recommendation that financial institutions treat AI as “virtual employees”, instead emphasizing that companies must realize that AI agents do not think like humans and may understand predefined goals in unexpected ways.

Related Reading:

Original source

IT之家 AI

Content notes

Original publication and rights belong to the source.

Machine translation · Refer to the original