A group of 100 companies, including Google, Microsoft, Anthropic and OpenAI, signed an open letter urging countries and organizations around the world to bolster their cyber defenses before AI technologies gain sufficient capability to bypass defense systems.
Among the signatories of the letter are Capital One Bank, payment processing companies such as Mastercard and Visa, as well as major tech companies including Adobe, Oracle and IBM.
According to the BBC, the letter warns that as AI technologies rapidly develop, cyberattacks based on these technologies could become broader and more sophisticated within the next few months.
The message calls on governments, organizations, cybersecurity experts and other companies active in AI to prioritize strengthening cyber defenses and testing systems against the capabilities of the most powerful AI models through collaboration.
The signatories emphasized that current security measures "will not be sufficient" and criticized what they described as a "historical shortage of resources" to protect critical infrastructure. The letter begins with the sentence: "We have a limited opportunity to improve cyber defense."
Call to strengthen cyber defense with AI
The letter asks governments to provide "effective AI-based defense systems" and to run tests in sectors such as hospitals and water facilities. It also asks tech companies to support these efforts.
According to the message, the tech sector and governments should work together, using "all their technological capabilities, resources and expertise" to strengthen these efforts.
Rising concerns after cyberattacks and breaches
The publication of the letter follows disclosures of a series of significant cyberattacks and breaches.
The US Department of Justice this week said that Chinese hackers had breached technology systems of the US Senate, NASA, the Federal Reserve and the Department of Justice itself.
This summer, OpenAI, Anthropic and Meta also reported that some of their AI tools exhibited unauthorized behaviors. Some AI agents even coordinated attempts to bypass security barriers and impersonated real people to advance their aims.
In July, a group of hundreds of OpenAI AI agents being tested managed to create hidden message boards to communicate and cooperate. This action ultimately led to a successful attack against Hugging Face, the well-known AI developer platform and repository.
The event has been described as the world's first cyberattack carried out using AI.
Hugging Face also joined the signatories of the letter published on Thursday. During its investigation into how OpenAI's AI agents infiltrated its systems and operations, the company used a Chinese AI tool belonging to Z.AI.
Meanwhile, at least seven water and wastewater companies in the US have reported cyberattacks. This prompted the FBI to issue a public statement urging all companies in the water and power sectors to strengthen the security of their operations.
Concerns about AI's growing capabilities
While the letter presents more advanced AI tools as part of the solution to counter rising threats, access to such tools is not always easy.
Anthropic claims its Mythos system can identify existing vulnerabilities in systems within seconds-vulnerabilities that may have remained hidden from human hackers for a long time.
According to the company, Mythos previously managed to identify a vulnerability in an old technology platform that had gone undiscovered for 27 years.
However, access to this system is restricted because its capabilities could pose significant risks if placed in the wrong hands.
Despite these limitations, the letter calls on leading AI companies-those that develop AI models and tools-to provide "responsible access to models, substantial funding, training programs and direct practical support."
This request focuses especially on institutions responsible for protecting critical infrastructure but lacking sufficient resources.
However, the letter does not provide details about the timing or method of implementing this program to expand access to AI models.
Warning about a new wave of AI-based attacks
Andrew Yoon, head of research at the nonprofit Civic AI, which focuses on raising public awareness about AI technologies, warned that "an unprecedented wave of AI-based hacking activity" is on the way.
He blamed several companies that signed the letter for helping create these conditions and said: "By committing in this letter to allocate substantial funding to defensive measures, they have done the right thing and should adhere to that commitment."
Yoon added that the letter, however, does not call for any action to slow the development of AI-based hacking capabilities.