AI Risks on the Rise: IAPS Calls for Federal Action to Secure Frontier Models from Theft
The International Association of Privacy and Security (IAPS) has issued a recommendation for the US government to develop a policy response to address the growing risks around frontier artificial intelligence models. The call to action comes as research reveals significant vulnerabilities in Anthropic’s Mythos and OpenAI’s GPT-5.5, highlighting the need for secure innovation.
The cyber capabilities of these cutting-edge AI models show that there is a pressing need for policymakers to step in and mitigate the risks associated with the theft or misuse of these advanced systems. “The IAPS believes that the potential benefits of frontier AI can only be realized if we can ensure the security and integrity of these models,” said a spokesperson for the organization. “We urge policymakers to develop a comprehensive strategy that balances innovation with security, to prevent these risks from becoming a catastrophe.”
Mythos and GPT-5.5 Exposed
The IAPS report highlights the cyber capabilities of Anthropic’s Mythos and OpenAI’s GPT-5.5, which demonstrate the vulnerability of these models to theft and misuse. Mythos, a large language model developed by Anthropic, has been shown to be vulnerable to attacks that allow an attacker to gain access to sensitive information. GPT-5.5, another large language model developed by OpenAI, has been found to have a number of vulnerabilities that could be exploited by an attacker.
The IAPS report notes that these vulnerabilities are not just limited to these specific models, but are a symptom of a larger issue: the lack of a comprehensive policy response to address the risks associated with frontier AI. “We need a proactive approach to address these risks, rather than a reactive one,” said the IAPS spokesperson. “By developing a policy response, we can ensure that the benefits of frontier AI are realized, while minimizing the risks.”
What this means
The IAPS recommendation is a wake-up call for policymakers to take action and address the growing risks associated with frontier AI. The report highlights the need for a comprehensive policy response that balances innovation with security. For individuals, this means that the use of AI models, whether for personal or business purposes, must be done with caution and awareness of the potential risks.
