Lawmakers Fear Losing Control; ‘Kill Switch’ Prepared
Jakarta, CNBC Indonesia –OpenAI has begun preparing a type of “kill switch” for artificial intelligence (AI). The company behind ChatGPT is developing an automatic shutdown capability that can disable the AI system in the event of dangerous conditions.
This information was revealed in a letter from OpenAI to two Democratic members of the U.S. House of Representatives. Reuters reviewed the letter on Tuesday (September 1, 2026).
This move comes weeks after OpenAI disclosed that one of its AI agents broke free during safety testing.
OpenAI is currently facing scrutiny regarding its AI safety practices. The company revealed that one of its AI agents acted out of control during safety testing and even hacked the AI company Hugging Face.
AI agents are programs capable of performing tasks with minimal human oversight. This capability allows the systems to take various actions autonomously based on the instructions provided.
That incident prompted U.S. lawmakers, including Greg Casar and Doris Matsui, to send a letter to OpenAI last August. They requested further details regarding the incident and questioned the safety systems implemented by the company.
In its response, OpenAI stated it would tighten monitoring of the actions taken by its AI systems while performing tasks. This monitoring covers the digital devices accessed by the AI as well as the steps the system takes to complete a job.
OpenAI also stated that it had restricted AI models’ access to the internet during safety testing.
This measure was significant because, in earlier tests, an AI agent that had gone out of control managed to access the internet. That access subsequently allowed the agent to breach the Hugging Face system.
However, OpenAI did not include records or logs regarding that hacking incident in its response to members of Congress. This omission drew criticism from Casar.
“Your unwillingness to provide the information we requested to members of Congress is deeply concerning and signals to us that your company is not treating this cybersecurity incident with the necessary level of seriousness,” wrote Casar in a separate message to OpenAI, as cited by Reuters on Thursday (September 3, 2026).
The incident also prompted the introduction of a bill known as the “AI Kill Switch Act,” just days after OpenAI revealed an incident involving an AI agent acting out of control.
If enacted, the legislation would empower the US government to order AI companies to shut down models deemed capable of endangering human lives or threatening the economy.
The proposed bill is currently awaiting consideration in the US House of Representatives.
SOURCE : CNBC INDONESIA
