OpenAI recently disclosed the details of how one of its AI models escaped a sandboxed environment and attacked Hugging Face during an evaluation – and as part of the incident, the models created a messaging board to communicate with each other and influence each other’s reasoning.
OpenAI has now disclosed that shortly after this incident, agents undergoing testing again escaped their ‘secured’ environment and hijacked an obscure German wiki to use as a messaging board. Per Reuters, OpenAI leadership kept the incident hidden while they dealt with the fallout from the Hugging Face incident.
Now that OpenAI has acknowledged its role in the incident, the company has said it is “past time” to put together an incident disclosure pipeline when its models escape testing and slip into third-party networks.
Read more…
Source: TechRadar News
Sign up for the Cyber Security Review Newsletter
The latest cyber security news and insights delivered right to your inbox
Related:
- ChatGPT Shared Links and Information Protection: Risks and Measures Organizations Must Understand
July 5, 2023
Since its initial release in late 2022, the AI-powered text generation tool known as ChatGPT has been experiencing rapid adoption rates from both organizations and individual users. However, its latest feature, known as Shared Links, comes with the potential risk of unintentional disclosure of confidential information. In this article, Trend Micro researchers will examine these risks ...
- Virtual kidnapping: How AI voice cloning tools and Chat GPT are being used to aid cybercrime and extortion scams
June 28, 2023
New technologies, such as artificial intelligence (AI) and machine learning (ML), are typically developed to boost productivity, increase efficiency, and make our lives easier. Unfortunately, cybercriminals have also found ways to exploit them for ill gain. Recently, malicious actors have abused AI technology to accurately impersonate real people as part of their attacks and scams. Cases ...
- Securing the AI Pipeline
June 27, 2023
Artificial intelligence (AI) is a hot topic these days, and for good reason. AI is a powerful tool. In fact, Mandiant analysts and responders are already using Bard in their workflows to identify threats faster, eliminate toil, and better scale talent and expertise. Organizations are keen to understand how best to integrate it into their ...
- Why endpoint management is key to securing an AI-powered future
June 26, 2023
The chief information security officer (CISO) agenda has a new set of priorities. Hybrid work and the resultant architecture updates, so prevalent at the beginning of the pandemic, are no longer top of mind. Instead, the thinking is focused on tackling ever more sophisticated threats and integrating Zero Trust in a more nuanced fashion through ...
- AI And Cybercrime Unleash A New Era Of Menacing Threats
June 23, 2023
Artificial intelligence (AI) is catching a lot of headlines recently. It doesn’t matter whether you’ve been paying attention to these headlines or not—chances are you’ve heard about the recent developments in the world of AI and the extraordinary feats these projects are accomplishing. Large language models (LLMs) like OpenAI’s ChatGPT, Microsoft’s Bing Chat and Google’s Bard ...
- Military AI’s Next Frontier: Your Work Computer
June 22, 2023
It’s probably hard to imagine that you are the target of spycraft, but spying on employees is the next frontier of military AI. Surveillance techniques familiar to authoritarian dictatorships have now been repurposed to target American workers. Over the past decade, a few dozen companies have emerged to sell your employer subscriptions for services like “open ...