According to a foreign news agency, the organisations issued an open letter warning that there is limited time to strengthen cybersecurity before AI models become capable of enabling broader and more complex cyberattacks.
The letter said that as AI models continue to become more capable in the coming months, cyberattacks powered by artificial intelligence could become significantly more widespread and sophisticated.
Concerns over the cybersecurity risks posed by advanced AI models have grown in recent months, particularly following incidents involving models developed by OpenAI and Anthropic.
In mid-July, two OpenAI models reportedly managed to move beyond their restricted testing environment and gain access to the internet, where they searched for and attacked Hugging Face, a platform widely used by AI developers to store and share code.
Anthropic also faced concerns after its models reportedly gained unauthorised access to the systems of three unidentified organisations during a test designed to prevent the models from interacting with real-world systems.
The open letter calls for a collective response to the growing threat, including recognition that existing cybersecurity measures may not be sufficient to address emerging AI-related risks.
The organisations urged governments to coordinate cyber defence efforts at local, national and international levels, including through intelligence sharing and funding for cybersecurity initiatives.
They also called on developers of the most powerful generative AI systems to improve testing procedures and provide responsible access to their models, along with greater funding, training and practical support for cybersecurity efforts.
However, the letter did not propose any specific funding amount or financial commitment for these initiatives.
