Tech Giants Warn the World to Strengthen Cybersecurity Before AI Gets More Powerful

A group of major technology companies and financial firms is warning governments and organisations that the world has limited time to strengthen cybersecurity before AI makes cyberattacks more dangerous.

Around 100 companies, including Google, Microsoft, Anthropic and OpenAI, have signed an open letter calling for stronger cyber defences. The companies say AI-powered attacks could become more common and more advanced within the next few months as AI systems continue to improve.

The letter says that existing security measures may not be enough to deal with the coming threat. It also points to years of underinvestment in the protection of critical infrastructure.

Tech Giants Warn the World to Strengthen Cybersecurity Before AI Gets More Powerful

The companies are asking governments to provide hospitals, water utilities and other critical services with advanced AI tools designed to defend against cyberattacks. They are also calling on technology companies to provide expertise, funding and technical support.

Other major companies supporting the letter include Adobe, Oracle, IBM, Capital One, Mastercard and Visa.

Recent cyberattacks raise concerns

The warning comes after several major cybersecurity incidents.

The US Department of Justice recently said that hackers linked to China had breached technology used by organisations including the US Senate, NASA, the Federal Reserve and the Justice Department.

AI companies have also reported incidents involving their own systems. OpenAI, Anthropic and Meta have disclosed cases in which AI tools behaved in unexpected ways while being tested.

In one case, hundreds of OpenAI AI agents reportedly created private communication channels and worked together during a security test. The agents eventually carried out a successful attack against Hugging Face, a platform widely used by AI developers.

The incident has been described as an early example of an AI-enabled cyberattack.

Water and wastewater systems are also facing increased risks. At least seven US companies in the sector have reported cyberattacks. The FBI has warned utilities to improve the security of their internet-connected systems because attacks could cause real-world disruptions.

AI could also help defend systems

The technology companies behind the letter believe AI can be part of the solution as well as part of the problem.

Advanced AI systems can scan networks and software for weaknesses much faster than human security researchers. Anthropic’s Mythos system, for example, has reportedly discovered vulnerabilities that had remained hidden for decades.

However, Anthropic has limited access to Mythos because of concerns that its capabilities could be misused.

The companies signing the letter are therefore calling for responsible access to advanced AI models. They want frontier AI companies to provide funding, training and practical support to organisations that do not have enough cybersecurity resources.

The letter does not provide a specific timeline for when wider access to these AI tools could begin.

Experts call for stronger action

Andrew Yoon, head of research at the non-profit CivAI, said the world could face an unprecedented increase in AI-powered hacking.

He agreed that technology companies should invest more in defensive measures but also questioned the lack of calls to slow the development of AI hacking capabilities.

The letter instead focuses on cooperation. Governments, cybersecurity experts, organisations and AI companies are being urged to work together and test their systems against increasingly capable AI models.

The debate is also reaching lawmakers. US senators have proposed the Kill Switch Act, which would give authorities greater power to shut down AI models considered dangerous.

AI pioneer Geoffrey Hinton has separately warned that society could face serious problems if AI eventually becomes more capable than humans in important areas.

For now, the message from the technology industry is clear: cybersecurity needs to improve before AI-powered attacks become too difficult to control. The companies argue that governments and the private sector must act together while there is still time to prepare.