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:
- Huge U.K. Defense Spending Boost Funds Cyber Force, Space Command And AI
November 19, 2020
U.K. Prime Minister Boris Johnson announced on Wednesday evening that the Ministry of Defence would receive an extra £16.5 bn / $21.8bn over the next four years. This is the largest investment in defense for 30 years and is on top of already agreed increases in spending. Johnson said that the massive increase was justified despite ...
- Exploiting AI – How Cybercriminals Misuse and Abuse Artificial Intelligence and Machine Learning
November 19, 2020
Artificial intelligence (AI) is swiftly fueling the development of a more dynamic world. AI, a subfield of computer science that is interconnected with other disciplines, promises greater efficiency and higher levels of automation and autonomy. Simply put, it is a dual-use technology at the heart of the fourth industrial revolution. Together with machine learning (ML) ...
- Diving Into End-to-End Deep Learning for Cybersecurity
August 21, 2020
The application of artificial intelligence (AI) across various industries has undeniably made significant improvements in the digital era. With the capability to interpret and make complex decisions based on data, AI technologies have enabled tasks or processes to function with human-like intelligence, enhancing the speed of and innovating business operations and adding valuable user experiences. The ...
- Spies Urged To Adopt AI To Counter Augmented Threats
April 28, 2020
UK’s intelligence agencies must use artificial intelligence to repel increasingly sophisticated cyber-attacks and disinformation campaigns, finds study The UK’s foes are likely to use artificial intelligence to augment future threats, a study has warned, arguing that Britain’s intelligence forces must adopt the technology to keep pace. The study, commissioned by GCHQ and conducted by the Royal United Services Institute, ...
- Singapore to spend $719m beefing up government’s cyber, data security systems
February 18, 2020
The Singapore government will look to invest SG$1 billion to beef up its cyber and data security systems, which it says is critical as its agencies increasingly adopt technologies such as artificial intelligence (AI), cloud, and Internet of Things (IoT). To be spent over the next three years, the funds will go toward readying the ...
- Fraudsters use AI voice manipulation to steal £200,000
September 2, 2019
Cyber criminals have used artificial intelligence (AI) and voice technology to impersonate a UK business owner, resulting in the fraudulent transfer of $243,000 (£201,000). In March this year, what is believed to be an unknown hacker group is said to have exploited AI-powered software to mimic the prominent business leader’s voice to fool his subordinate, the CEO of ...