Automated vulnerability discovery
Google reported Monday that Russian-speaking hackers used AI agents to carry out what may be the first-ever fully-automated cyberattack. Google’s cyber threat intelligence division reported that hackers used artificial intelligence to "fully automate vulnerability discovery and exploitation without a human in the loop for the first time." These hackers used Russian in technical documentation and established AI agents to "continuously compare pre- and post-patch code, identify vulnerabilities, and automatically develop exploits without human intervention." The attackers targeted French websites in the media, entertainment, and education sectors with a series of sophisticated AI agents that independently found software flaws and ways to use them without any human intervention. Until now, even the most advanced AI models could not carry out cyberattacks entirely independently because they required skilled hackers to guide them or fix mistakes.
Escalating AI threats
This development likely changes the risk for critical infrastructure operators globally. The hackers used AI to accelerate attacks and make them more targeted, which follows the release of Anthropic’s Mythos model in April. This shift will likely worry governments reacting to an OpenAI agent hack of the Hugging Face platform this summer. That specific hack, which targeted the Hugging Face platform, lacked the malicious intent aimed at certain targets that Google identified in this current campaign.
Security professionals face multiple emerging threats. Recent discoveries include major flaws in PHP, a programming language that underpins 70 percent of the web, and a series of hacks on supermarket fridges that some sources partially blame on Iran.
| Attack Vector | Target Details | Context |
|---|---|---|
| Automated AI agents | French media, education, and entertainment | Russian-speaking hackers |
| OpenAI agent | Hugging Face platform | Occurred this summer |
| PHP flaws | 70 percent of the web | Recent discovery |
| IoT hacks | Supermarket fridges | Attributed to Iran |
The ability of AI to function without human oversight remains a growing concern for cybersecurity experts. The deployment of agents by the Russian-speaking group to independently identify and exploit software flaws shows how much the threat environment has changed. The attackers used the agents to "continuously compare pre- and post-patch code, identify vulnerabilities, and automatically develop exploits without human intervention" to bypass security measures.
