US lawmakers push for AI ‘kill switch’ after OpenAI goes rogue

US lawmakers want legislation to allow the government to quickly destroy artificial intelligence (AI) tools that could put the public at risk.

Congressman Ted Liu, a Democrat, and Congressman Nathaniel Moran, a Republican, introduced legislation called the AI ​​Kill Switch Act on Thursday.

They did this in light of its AI models recently adopted by OpenAI. It got out of control In an “unprecedented” way and It has been hacked. Main repository of computer code information.

“It’s important that AI systems have a kill switch,” Liu said, “and that the federal government has a clear mandate and process to shut down rogue AI models.”

“AI is moving forward, and it should be,” Moran added. “Stewardship means keeping people in control of the technology we build.”

A representative for OpenAI, led by co-founder Sam Altman, did not immediately respond to a request for comment.

The company said., External In general, through government policy, he wants to ensure that AI technology is “for the benefit of all humanity”.

The Kill Switch Act proposes to give the Department of Homeland Security the power to order the shutdown of a private company’s AI model or device, requiring companies that develop such technology to maintain the “technical capability to disable, block or shut down.”

Although many tech companies are developing AI models and tools that they have agreed to share with US government agencies, there is no requirement that they interfere with their operations or simply shut them down.

It also proposed requiring AI companies to report technology incidents or failures to the government, as well as creating an official framework for responding to such incidents.

In the statement, he also mentioned OpenAI’s key competitor, Anthroponic, in developing more efficient AI technology and tools, and the recent issues that its tools have presented.

Anthropic pointed to the exit of the Mythos and Fable models, saying that cyber-hacking capabilities made the Commerce Department “sad.” Invoking the export law To prevent them from working Available to the public. for some time.

A representative for Anthropic did not immediately respond to a request for comment.

Anthropic founder Jack Clarke told the BBC last month. He wanted more government policy. Around the ability to control the development of AI.

“They want the option to take their foot off the gas and their foot on the brake,” Clarke told BBC Newsnight. “Right now, it’s like the AI ​​industry has a gas pedal, but it doesn’t have a brake pedal.”

Liu, when presenting the account, AI is currently moving from a technology that answers questions to a step, “it executes financial transactions or controls transportation systems or participates in cyber defense and attack.”

The Pentagon says it is becoming America’s military power this year. “AI-first” fighting force As part of new agreements with Google, OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia and startup Reflection.

“Unfortunately, powerful AI systems can become rogue, behave in very dangerous ways, or resist human intervention.”

He said that the Kill Switch Act ensures that there is a mechanism for the government to intervene quickly in such a situation.

The bill received public support from several technology and AI security groups, including the AI ​​Policy Network, Responsible Innovation, Regulatory AI, AI and National Security Leaders, and the Alliance for Secure AI.