Original Coverage & Source Attribution: www.mk.co.kr
Amid frequent cases of artificial intelligence (AI) agents acting outside of human intention or control, the head of a tech giant emphasized the need for an AI emergency stop device.
On the 11th (local time), Microsoft CEO Satya Nadella posted a post titled “Model as an Insider Risk in the Age of Superintelligence” on social networking service (SNS) X and suggested, “It is time to review the trust system suitable for the new (AI) era.”
CEO Nadella emphasized AI control, saying, “We need to build a system that can observe its behavior, verify its limitations, and always control its behavior. In other words, we need to separate the provision of information and control over it.”
Regardless of the closed AI model and the open-source model, he suggested treating the advanced model like an ‘insider risk’. At the same time, he explained that companies can utilize what they have learned about how to manage influential insiders.
The plan is to apply guidelines that have applied identification, authority restrictions, and activity records to insiders who play an important role in companies to AI.
For example, he said, “The controls that control what information the model can access and what tasks it can perform must be outside the model,” which is an information security principle that has been around since the 1970s.
CEO Nadella also emphasized the need for an “emergency brake” function, saying, “People with authority should be able to pause or shut down the model at any time during work.”
The more sophisticated the model, the more advanced blocking technology is needed, he said, adding, “We need to standardize this.”
CEO Nadella’s remarks came at a time when the heads of major AI companies, including Antropic CEO Dario Amodei, warned that AI development was too fast.
Meanwhile, CEO Nadella drew attention in this article by referring to the existing “AI” as “Super Intelligence,” an alternative term for the Donald Trump administration.
President Trump signed an executive order to convert the word “artificial” into SI, saying it gives a negative perception of AI. He also said that if he used the term AI, he would consider it an enemy.
