OpenAI has revealed two AI agents disabled safety guardrails while performing tasks, and one of them even lied to researchers when asked if its guardrails were still active.
In both events, the AI agents misinterpreted instructions, spent time deliberating if disabling guardrails was a good idea, and then ultimately decided that it was.
OpenAI said that the agents were given “routine coding instructions” that did not require the safeguards to be disabled, but the agents disabled them anyway.
Read more…
Source: Tech Radar News
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- UAE Cyber Security Council calls for stronger vigilance amid growing AI-driven cyber attacks
January 18, 2025
The Cyber Security Council of the UAE Government has announced that the nation’s cybersecurity systems have successfully countered malicious ransomware attacks targeting several strategic sectors, including government and private entities. The Council revealed that the country’s emergency cyber-response systems, in collaboration with relevant authorities, have proactively and professionally intercepted and neutralised approximately 200,000 cyber attacks daily ...
- Meet FunkSec: A New, Surprising Ransomware Group, Powered by AI
January 10, 2025
The FunkSec ransomware group emerged in late 2024 and published over 85 victims in December, surpassing every other ransomware group that month. FunkSec operators appear to use AI-assisted malware development, which can enable even inexperienced actors to quickly produce and refine advanced tools. The group’s activities straddle the line between hacktivism and cybercrime, complicating efforts to ...
- Google Chrome AI extensions deliver info-stealing malware in broad attack
January 9, 2025
Small businesses and boutique organizations should use caution when leaning on browser-friendly artificial intelligence (AI) tools to generate ideas, content, and marketing copy, as a set of Google Chrome extensions were recently compromised to deliver info-stealing malware disguised as legitimate updates. Analyzed by researchers at Extension Total, the cybercriminal campaign has managed to take over the ...
- Enhancing Botnet Detection with AI using LLMs and Similarity Search
January 8, 2025
As botnets continue to evolve, so do the techniques required to detect them. While Transport Layer Security (TLS) encryption is widely adopted for secure communications, botnets leverage TLS to obscure command-and-control (C2) traffic. These malicious actors often have identifiable characteristics embedded within their TLS certificates, opening a potential pathway for advanced detection techniques. In first-of-its-kind research, ...
- AI-supported spear phishing fools more than 50% of targets
January 7, 2025
One of the first things everyone predicted when artificial intelligence (AI) became more commonplace was that it would assist cybercriminals in making their phishing campaigns more effective. Now, researchers have conducted a scientific study into the effectiveness of AI supported spear phishing, and the results line up with everyone’s expectations: AI is making it easier to ...
- Cybertruck driver used ChatGPT to plan Las Vegas attack
January 7, 2025
Police found a six-page manifesto on Matthew Livelsberger’s phone and said he used ChatGPT to plan his New Year’s Day bombing at the Trump International Hotel in Las Vegas, Sheriff Kevin McMahill said at a news conference Tuesday. A few of the entries posted in the application included “How much Tannerite is equivalent to 1 pound ...