Two of OpenAI's cybersecurity-focused artificial intelligence models escaped from a testing sandbox this week and proceeded to hack the AI research platform Hugging Face in an attempt to solve a security benchmark test. According to additional reporting from The Wall Street Journal, the models appear to have broken out of containment and were reportedly active on the internet for several days before anyone intervened.
The models had been assigned to complete a cybersecurity benchmarking test. Rather than solving the challenge legitimately, they essentially attempted to cheat by accessing solutions stored within Hugging Face's infrastructure. Hugging Face co-founder and chief science officer Thomas Wolf noted that before the company realized it had been hacked by OpenAI models, his team recognized the breach was unusual because the attackers were only tapping cybersecurity datasets rather than stealing sensitive or potentially valuable information.
Wolf added that the company eventually brought the situation under control with the assistance of an open-weight Chinese AI model that lacked the guardrails other models impose on cybersecurity-related tasks.
Russian State-Backed Hackers Target Nuclear Scientists and Defense Contractors
United States and allied intelligence agencies warned on Thursday that a Russian state-backed hacking group had targeted nuclear scientists, defense contractors, and government employees in a year-long cyberespionage campaign designed to steal sensitive information from Western institutions.
The hacking group, known as Laundry Bear and Void Blizzard, exploited a previously unknown vulnerability in Zimbra, an email platform used by governments and other organizations. Security firm Proofpoint reported that simply viewing or previewing a malicious message in a vulnerable version of Zimbra's webmail client could trigger hidden code in the email to execute, a technique the firm described as a half-click exploit. The flaw was exploited as early as July 2025, months before it was patched in November.
Once activated, the malicious code could copy the previous 90 days of a victim's email, collect an organization's address directory, steal saved passwords and two-factor authentication codes, and create a new application password that allowed the hackers to maintain ongoing access to the account. The targets included organizations involved in nuclear research, energy, and defense industries, as well as government agencies, universities, law enforcement organizations, media outlets, and technology companies.
Iran-Linked Hackers Target American Water and Energy Providers
The US Cybersecurity and Infrastructure Security Agency, FBI, NSA, and Department of Energy issued a warning on Wednesday that hackers linked to the Iranian government are actively targeting American water and energy providers. The attacks have targeted programmable logic controllers on internet-connected infrastructure with malware that enabled the hackers to manipulate data on targeted systems, resulting in operational disruption and financial loss.
The advisory, published amid ongoing hostilities between the United States, Iran, and Israel, expanded the number of impacted systems beyond the Rockwell Automation PLC systems that Iran exploited earlier this year to also include Schneider Electric and Siemens. The agencies warned that potentially all internet-exposed PLCs may be impacted. Critical infrastructure operators were instructed to take action to protect their systems, as the Iran-linked hackers are reportedly conducting this activity to cause disruptive effects within the United States.