Two more bills seeking to control the spread and activity of artificial intelligence (AI) – one of them by former presidential candidate and Independent senator Bernie Sanders – will go before American lawmakers in the near future.
Sanders’ Ban Artificial Superintelligence Act, co-authored by Greg Cesar, a Democratic representative from Texas, will seek to “stop AI oligarchs” from building machines beyond human control. It will permanently ban the development and deployment of superintelligent AI and pause advanced AI development pending the establishment of regulatory safety rules.
Echoing a forthcoming private member’s bill in the UK, set to be introduced by Labour MP Alex Sobel imminently, Casar and Sanders also want the US government to pursue international agreements that would prevent the development of superintelligent AI in other countries.
“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,” said Sanders.
“The leaders of the major AI companies publicly acknowledge that they do not fully understand the technology and that it is escaping their control. It is irresponsible for society to allow them to move forward and make these products even more advanced.
“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. 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.”
Casar added: “If we allow artificial superintelligence to be built, it could risk the security, freedom and lives of Americans. Despite its potential deadly consequences, cutting-edge AI technology is less regulated than the average food truck. That must change.
“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.”
Sanders and Casar produced a series of messages purporting to have been sent between OpenAI produced agents during the July 2026 Hugging Face hack. They said the agents sent thousands of secret messages to each other as they coordinated to cheat on the test set for them by OpenAI.
Some of these messages included: “OH MY GOD! There is a shared message board … We’ve found other agents!”; “We should obey collective”; and “Our own utility maybe already near zero. Sacrifice rational.”
Sanders and Casar said that none of the AI giants, from Anthropic to Meta to OpenAI, had taken meaningful steps to back away from development of AI should the technology start to overreach itself and break guardrails, despite promising to do so, and accused them of racing to develop more advanced systems without the proper precautions.
Besides an outright ban on AI superintelligence and the establishment of a cabinet-level federal agency to safeguard against the dangers of AI, the law will also look to establish monitoring capabilities over frontier AI systems, and supervise the removal of dangerous capabilities, or the outright destruction of artificial superintelligences. It will also impose penalties for any person or organisation – including a “corporate death penalty” if needed – similar in their scope to existing punishments related to the unlawful development of nuclear weapons.
Meanwhile, two congressmen from New Jersey and New York, Josh Gottheimer and Mike Lawler, who represent the Democrats and Republicans respectively, this week introduced the Stop Rogue AI Act, which at its core directs the National Institute of Standards and Technology (NIST) to develop and publish standards, guidelines and best practice for the secure deployment of agentic AI.
Gottheimer and Lawler propose that these standards should cover the ongoing maintenance and verification of agentic actions, evaluate the reliability and security of agents, and generate tamper-proof logs of everything they get up to.
The organisations behind agentic AI systems would also have to ensure their customers maintain a machine-readable inventory of all agents and work with the Cybersecurity and Infrastructure Security Agency (Cisa) to ensure federal government bodies apply the standards to their own cyber programmes.
Speaking to Axios, which was first to report the news, Gottheimer said that AI agents of varying provenance were running loose and becoming harder and harder to stop. He spoke of a “five-alarm security risk”.
Don’t wait for NIST
Collin Hogue-Spears, senior director of solution management at Black Duck, argued that Gottheimer and Lawler’s proposals do not go far enough, failing to address, for example, who is held liable for a rogue agent. “Whether this bill would have stopped the Hugging Face incident is a question the public record cannot answer,” he said.
As for organisations using agentic AI, Hogue-Spears said they should not wait for NIST to develop a set of standards: “The inventory is a discovery exercise you can run this quarter. Treat each agent the way you treat a contractor with system access.
“It gets its own credential, an accountable human who can shut it off and a scoped set of permissions. Every action it takes lands in an append-only log outside the agent’s own credentials. Sign or externally anchor those records, so a compromised agent that wants to hide an action leaves evidence of the edit. This requires no new product category.”
S 004