Nadella calls for emergency brake for advanced artificial intelligence

Microsoft CEO Satya Nadella has stated that advanced artificial intelligence models need an emergency stop mechanism. This is reported by Business Standard, citing statements by Nadella himself.

This news currently comes from a single source; no independent confirmation is available at the time of publication regarding the exact context in which the statement was made or any technical details of the mechanism proposed by Nadella.

The statement comes after the disclosure of a security breach by Anthropic, an episode that has reignited public debate over the safety of the most advanced artificial intelligence systems. A group called Super Intelligence Force has asked companies in the sector to promptly disclose incidents involving their models.

“Companies must immediately disclose incidents involving their models” reads the group’s statement, reported by Business Standard.

The available sources do not specify the exact nature of the breach suffered by Anthropic, nor whether the company has already adopted corrective measures at the time this news was published. It remains the case, however, that two different players in the sector — the CEO of the leading company investing in artificial intelligence and an independent group of experts — have, in the same context, converged in calling for faster and more transparent control tools for advanced models.

The issue raised by these statements concerns the speed with which companies communicate security incidents once discovered, rather than the existence of internal control mechanisms as such. It is on this point — the timeliness of disclosure, not its existence in absolute terms — that the request from the Super Intelligence Force group reported by Business Standard focuses.

It remains to be seen whether the two requests — Nadella’s for a technical emergency stop and the group of experts’ for immediate disclosure — will find concrete follow-through from companies in the sector, or whether they will remain isolated statements without a shared mechanism. At present, none of the available sources indicates timelines or methods for the practical adoption of either of the two tools requested.