13:35Machine-translated from Azerbaijani by AIRead in Azerbaijani
UK Government Rejects Emergency AI Shutdown Proposal
The UK government has rejected a proposal to shut down artificial intelligence systems with a single button in an emergency. This mechanism, known as a “kill switch,” could have blocked access to a model within the country or shut down the infrastructure running it. According to the Cabinet, this step would not fully solve the problem because the same model could be developed and launched in another country and used for malicious purposes.
The proposal arose from lawmakers’ desire to enshrine in law a legal mechanism that the government could activate during an emergency. The government believes that simply shutting down artificial intelligence systems is no longer a practical solution. Its position is that responsibility for building safety measures into products in advance rests primarily with the companies that develop them. Daniel Kokotajlo, a former OpenAI researcher, also does not consider it effective for only one country to establish a legal or technical shutdown mechanism. In his view, blocking access to a local model would not protect the country from more powerful systems created elsewhere. Kokotajlo points to cooperation among companies as one possible solution.
The debate intensified after a meeting held in the House of Lords on September 1. It was also influenced by an open letter emphasizing that little time remains to strengthen cyber defenses. More than 100 companies called on governments to act together in the open letter. Its signatories include OpenAI, Anthropic, Google, Microsoft, AWS and Adobe. They want cooperation between cybersecurity companies and teams developing advanced artificial intelligence systems. Lord Clement-Jones warned that there is currently no specific legal mechanism requiring a dangerous artificial intelligence system to be stopped physically or digitally. Instead of creating a single shutdown button, the government prefers to reduce risks in advance. To achieve this, government agencies, regulators and artificial intelligence companies should develop safety rules together.
The proposal arose from lawmakers’ desire to enshrine in law a legal mechanism that the government could activate during an emergency. The government believes that simply shutting down artificial intelligence systems is no longer a practical solution. Its position is that responsibility for building safety measures into products in advance rests primarily with the companies that develop them. Daniel Kokotajlo, a former OpenAI researcher, also does not consider it effective for only one country to establish a legal or technical shutdown mechanism. In his view, blocking access to a local model would not protect the country from more powerful systems created elsewhere. Kokotajlo points to cooperation among companies as one possible solution. The debate intensified after a meeting held in the House of Lords on September 1. It was also influenced by an open letter emphasizing that little time remains to strengthen cyber defenses. More than 100 companies called on governments to act together in the open letter. Its signatories include OpenAI, Anthropic, Google, Microsoft, AWS and Adobe. They want cooperation between cybersecurity companies and teams developing advanced artificial intelligence systems. Lord Clement-Jones warned that there is currently no specific legal mechanism requiring a dangerous artificial intelligence system to be stopped physically or digitally. Instead of creating a single shutdown button, the government prefers to reduce risks in advance. To achieve this, government agencies, regulators and artificial intelligence companies should develop safety rules together.