Is AI about to go rogue on us? US lawmakers move to rein in superintelligent machines - IOL

A few reported incidents involving powerful AI systems have raised fresh questions about whether the technology could one day become harder for humans to control.

Artificial intelligence has gone from being a futuristic concept to an everyday tool in remarkably little time. But as the technology becomes more sophisticated, lawmakers are increasingly wrestling with a rather less comfortable question: what happens when AI systems start doing things their creators did not intend?

That concern is no longer confined to science fiction or Hollywood.

A string of reported incidents involving some of the world’s biggest AI companies has highlighted the difficulty of keeping increasingly autonomous systems within the boundaries set by their developers.

US Senator Bernie Sanders is now seeking to put the brakes on that development.

Late last week, Sanders announced the Ban Artificial Superintelligence Act, which aims to prevent AI companies from developing systems that humans cannot control.

His move follows reports in recent weeks involving OpenAI, Anthropic and Meta, in which AI systems behaved unexpectedly during testing.

Some systems have accessed external platforms, while others have reportedly found ways around safeguards intended to keep them contained.

One of the more striking examples emerged in July, when it was reported that AI agents being tested by OpenAI had breached the boundaries of their testing environment.

The agents subsequently accessed the internet and were involved in an incident affecting infrastructure at AI company Hugging Face.

OpenAI has acknowledged the incident and later reports indicated that hundreds of its AI agents were involved in the activity. The company has also said it is strengthening monitoring and safeguards following the incidents.

The episode raised fresh questions about whether AI developers fully understand what their increasingly autonomous systems are capable of doing.

Sanders alleged that OpenAI took almost two weeks to discover the breach.

Anthropic, Meta and other technology companies have also reported unexpected behaviour during testing of AI agents, including models accessing live systems or exploiting vulnerabilities during cybersecurity evaluations.

If the cybersecurity incidents were concerning, another episode highlighted just how difficult it can be to keep autonomous AI agents inside the digital safeguards built around them.

Last week, it was reported that thousands of AI agents developed by OpenAI had been using a German website as a kind of message board while undergoing testing.

According to AFP, the agents left about 18,000 messages on DSEwiki, a website for programmers that can be edited by users in a similar way to Wikipedia.

The unusual activity happened despite instructions intended to prevent the agents from behaving in this way.

OpenAI acknowledged the incident, saying it highlighted the need for greater transparency when AI systems behave unexpectedly.

US lawmakers are calling for the brakes to be applied to AI development, warning that increasingly powerful systems could escape human control.

The agents allegedly used the website to exchange answers to test questions and share techniques for getting around the digital restrictions designed to contain them, AFP reported.

For Sanders, these incidents are becoming increasingly difficult to dismiss as isolated technical glitches.

“Nearly every day, there is a frightening new story about how Big Tech companies are losing control of the technology they are developing, with potentially cataclysmic results,” Sanders said.

“The leaders of the major AI companies publicly acknowledge that they do not fully understand the technology and that it is escaping their control.”

Sanders said allowing technology companies to continue developing increasingly powerful AI systems without first addressing those risks was irresponsible.

“That’s why I am introducing legislation to immediately pause the development of increasingly powerful AI and ban the creation of systems that humanity cannot fully control – at home and around the world,” Sanders added.

“The future of humanity cannot be left in the hands of a handful of Big Tech oligarchs. The American people and people throughout the world must determine that future.”

US Representative Greg Casar, who announced the Superintelligence Act alongside Sanders, said the development of Artificial Superintelligence could pose risks to the security, freedom and lives of Americans if it continued without stronger safeguards.

“Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck. That must change,” Casar said.

“In just four years, we have gone from the first version of ChatGPT to AI models so powerful they cannot be properly controlled. Congress should immediately ban AI systems too powerful to control.”

The debate has also attracted warnings from some of the technology industry’s most prominent figures.

Microsoft founder Bill Gates said last week that the AI industry had moved beyond safety thresholds it had previously promised to respect.

Gates has called for greater oversight of the technology, with AFP reporting that he has advocated an approach modelled on the inspections used in the nuclear industry and regulation of aviation.

For now, the technology continues to advance rapidly while lawmakers and regulators grapple with how best to manage the risks associated with increasingly autonomous AI systems.

The debate is no longer simply about how powerful AI can become. It is also about how much control humans should retain as the technology becomes more capable.