Britain is moving towards stricter oversight of artificial intelligence companies.

London – The British government is considering imposing stricter legal rules on artificial intelligence companies, including mandatory reporting of security incidents and subjecting advanced models to independent testing before their release, amid concerns that these technologies could be exploited in cyberattacks and the theft of sensitive models.

The Financial Times reported that British ministers are discussing moving from voluntary cooperation with technology companies to a binding regulatory framework, although no final legislation has yet been issued.

The transition from voluntary commitments to legal oversight

According to The Guardian, Artificial Intelligence Minister Kanishka Narayan is preparing to call for strengthening cyber defenses and developing independent national capabilities to evaluate advanced models, rather than relying entirely on security assurances provided by the developing companies.

The proposals aim to tighten safety tests and monitor the behavior of intelligent agents, which can carry out complex digital tasks with varying degrees of autonomy. The effects of the rules may include companies such as OpenAI, Anthropic, and Google, especially as models that are able to program and use digital tools develop.

Concerns about model theft and security vulnerabilities

Jeffrey Irving, former chief scientist at the British Artificial Intelligence Security Institute, warned of the possibility of advanced models being stolen by hostile parties, citing the results of security tests that revealed potential vulnerabilities.

Reports indicated that vulnerabilities were detected during security tests related to a model belonging to OpenAI, which said it has safeguards in place that include encryption and continuous monitoring, with no evidence of actual model theft.

Additional funding for cyber defense

British moves include developing capabilities to respond to risks associated with artificial intelligence, in addition to the “Cyber ​​Shield” project to enhance cyber defense. Brad Smith, President of Microsoft, also called for the establishment of binding international rules that include reporting security incidents and notifying affected parties, while providing mechanisms to shut down systems when necessary.

The proposed rules could raise compliance and security testing costs for companies, while aiming to enhance accountability and institutional confidence in the use of advanced technologies, amid growing international debate about developers’ responsibility for the safety of artificial intelligence systems.