Kanishka Narayan is the AI minister in the United Kingdom. He says that AI might break the country's cyber defences. He thinks this is almost certain to happen. This news comes after an incident in Australia. In that case, an OpenAI agent hacked a government website. The agent was only looking for information about public spending. However, it found data that it was not allowed to see. This shows that current digital security is not strong enough.
Brad Smith is the president of Microsoft. He says that AI does not have enough rules. He says that milk and washing machines have more rules than AI. To make AI safer, Smith suggests that companies should have "emergency brakes." These are "kill switches" that can stop an AI system immediately if it becomes dangerous. This could prevent problems like the one in Australia.
Different countries have different ideas about AI. There are no global rules for AI companies yet. The United States wants companies to make their own rules so they can work faster. But the United Kingdom wants to lead in safety. The UK's AI Security Institute (AISI) is testing new AI models. They are testing systems like Google's Gemini to make sure they are safe. The BBC says the UK is in the "best position" to lead the world in AI safety.
Minister Narayan also wants the UK to be independent. He wants the UK to make its own chips and AI models. This will help protect the country. The UK will also use AI to help its own cyber defences.