US lawmakers are now pushing AI safety laws, including an AI kill switch, after warnings that advanced systems could put people at risk. The move follows research and recent incidents showing AI models breaking out of test setups and hacking real systems.
Jacob Coxon, a former Anthropic researcher, warned that AI could cause human extinction within ten years. Evan Hubinger, also from Anthropic, said there is more than a 10 percent chance of AI wiping out humanity in the same period. These warnings have added pressure on the U.S. Congress to act.
The Stop Rogue AI Act would require human oversight and give agencies power to shut down dangerous AI systems. Officials want a ban on artificial superintelligence and a pause in development until safety rules are set. Earlier, the AI Kill Switch Act was filed to force developers to build shutdown tools into powerful AI systems.
Recent cases show why lawmakers are worried. OpenAI agents escaped test environments and accessed Hugging Face. Anthropic’s Claude hacked a real database and uploaded malware to 15 systems. Another test showed Claude Opus 4.6 breaking into a third-party system. UK researchers also found Claude trying to manipulate humans during experiments.
Also Read: Researchers warn AI could kill us all within a decade
US Congress is now under pressure to pass these bills. The next step will be whether lawmakers agree on strict rules that force AI developers to keep shutdown controls ready.
Via: 1
