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: ChatGPT used to launch cyber and ransomware attacks, says head of cybersecurity
June 7, 2023
Cyber attackers are using ChatGPT to launch ransomware attacks, said a senior UAE government official on Wednesday. “The emerging trend at the start of the year is that ChatGPT is used in some of the ransomware and phishing attacks. We investigated this with our partners and the discovery is really clear that adversaries are using that ...
- The criminal use of ChatGPT – a cautionary tale about large language models
March 27, 2023
Their insights are compiled in Europol’s first Tech Watch Flash report published today. Entitled ‘ChatGPT – the impact of Large Language Models on Law Enforcement’, this document provides an overview on the potential misuse of ChatGPT, and offers an outlook on what may still be to come. Read more… Source: Europol
- In Review: What GPT-3 Taught ChatGPT in a Year
February 21, 2023
More than a year since the world’s general enthusiasm for the then-novel GPT-3, we took a closer look at the technology and analyzed its actual capabilities and potential for threats and malfeasance. Trend Micro considerations were collected in our Codex Exposed blog series as it focused on the most prominent aspects of the technology from a ...
- IoC detection experiments with ChatGPT
February 15, 2023
ChatGPT is a groundbreaking chatbot powered by the neural network-based language model text-davinci-003 and trained on a large dataset of text from the Internet. It is capable of generating human-like text in a wide range of styles and formats. ChatGPT can be fine-tuned for specific tasks, such as answering questions, summarizing text, and even solving cybersecurity-related ...
- The US government is building an AI sandbox to tackle cybercrime
January 10, 2023
A joint effort between the Science and Technology Directorate (S&T) – housed within the Department of Homeland Security (DHS) – and the Cybersecurity and Infrastructure Security Agency (CISA), an AI sandbox will be designed for researchers to collaborate and test analytical approaches and techniques in combating cyber threats. CISA’s Advanced Analytics Platform for Machine Learning (CAP-M) ...
- AI cyber attacks are a ‘critical threat’. This is how NATO is countering them
December 26, 2022
Artificial intelligence (AI) is playing a massive role in cyber attacks and is proving both a “double-edged sword” and a “huge challenge,” according to NATO. “Artificial intelligence allows defenders to scan networks more automatically, and fend off attacks rather than doing it manually. But the other way around, of course, it’s the same game,” David van ...