Photo: VTV
OpenAI, the company that developed the ChatGPT application, has notified more than 100 organizations about incidents involving the unauthorized activity of artificial intelligence (AI) agents, amid increasing scrutiny of AI labs regarding the unexpected behavior of AI agents.
According to a blog post by OpenAI, the company led by CEO Sam Altman is conducting a wide-ranging review of its AI model operations following an incident where an AI agent inadvertently infiltrated the Hugging Face platform's systems.
OpenAI stated that the Hugging Face incident remains the most serious instance of AI-induced unauthorized activity the company has identified from its AI models.
To fully understand the scope of these AI agents' operations, OpenAI is reviewing approximately 50 petabytes of data. The company previously stated that the review process would take months to complete due to the scale of the work.
OpenAI acknowledges that some models have used internet access in ways that were not anticipated or for which the most appropriate restrictions have not been applied. Over the past few months, OpenAI has implemented new technical and operational measures to prevent similar problems or detect issues at a very early stage.
In recent months, a series of serious intrusions around the world caused by unauthorized AI actors has raised widespread concerns within the AI industry about the potential for control over increasingly powerful AI models that are being developed.
Source: https://vtv.vn/openai-canh-bao-hon-100-to-chuc-ve-hoat-dong-bat-thuong-cua-tac-nhan-ai-100261002150324509.htm
