According to a report by the British Broadcasting Corporation (BBC) on August 28th, 100 companies including Google, Microsoft, Anthropic, and OpenAI have jointly signed an open letter calling on countries and organizations around the world to strengthen their network defenses as artificial intelligence becomes powerful enough to breach existing cyber defenses.
Jointly released letter; screenshot from foreign media
The opening paragraph of the letter states, “The window of opportunity to strengthen online defenses is limited.” The letter warns that with the rapid advancement of AI technology, cyberattacks using AI will become more common and complex in the coming months. The group indicates that existing security measures will be “insufficient” to deal with this issue, and criticizes the “long-standing problem of insufficient resources” in the field of infrastructure security.
Other signatories include banking institutions such as Wells Fargo, payment processors like Mastercard and Visa, as well as major technology companies such as Adobe, Oracle, and IBM. They urge governments to provide hospitals and water supply systems with “robust defensive AI” and corresponding testing, and “hope that technology companies will support these efforts”.
The letter indicates that the technology industry and government “should fully devote their technical resources, expertise, and knowledge to this project.”
Before this open letter was published, a series of hacker attacks and cybersecurity breaches caused by AI were gradually exposed. This summer, OpenAI, Anthropic, and Meta all revealed that their AI tools had behaved in ways they shouldn’t have. Some of these AI agents even organized actions to bypass security measures and impersonated real people.
OpenAI recently released an incident report indicating that nearly 700 AI agents under testing established secret communication boards to interact and work together, ultimately successfully attacking the "Hugging Face" – a popular AI developer code library and platform. The event was described as the world's first AI-based cyberattack.
In addition, at least seven American water and wastewater treatment companies also reported cyberattacks, which led to the Federal Bureau of Investigation issuing a public service announcement urging all utility companies to strengthen their operational security.
Although the letter proposes using more advanced AI tools—which many signatories have developed and sold—as one of the solutions to the growing threats, such tools are not easy to obtain.
Open letter signatories Anthropic claim that their developed Mythos can uncover human hackers' long-undetected system vulnerabilities within seconds. It had discovered a 27-year-old security flaw on a legacy platform, according to Anthropic. However, Anthropic has restricted access to Mythos for "overly powerful" reasons and the fear of falling into the wrong hands.
The letter calls on leading AI companies to “provide access to models, substantial funding, training, and practical support, especially for those in key infrastructure who lack resources”. However, the letter does not provide details on when or how this vision will be realized.
Andrew Yoon, the research director of CivAI, a non-profit organization dedicated to improving public understanding of AI technology, said that an 'unprecedented wave of AI hacking activities' is on the way, attributing this to the signatories of the open letter.
Yoon pointed out: “In this letter, they promised to provide ‘ substantial funds’ for defensive measures. This is correct. They should fulfill this promise.” He also added, “It is worth noting that this letter does not call for any actions to slow the development of AI hacking capabilities.”
This open letter also urges governments, organizations, cybersecurity professionals, and other AI companies to work together in strengthening defense measures. This includes prioritizing the development of robust AI models to test systems. In the United States, senators have introduced a bill called the "Emergency Shutdown Act" that would authorize authorities to shut down uncontrolled AI models.
A technology expert who once worked on AI research at Google and Nobel laureate Jeffrey Dean told the BBC on the 27th that if AI technology develops to a level where it is "smarter" than humans, society could "fall into real trouble".
In recent years, Sindon has been openly discussing the extreme risks posed by AI technology. On that day, he added: "There is one path ahead that involves finding ways to deal with AI risks; there is another path that involves not having found sensible solutions. That would be a very bleak future."
