September 12, 2024 at 08:24AM Ireland’s Data Protection Commission (DPC) has initiated a cross-border inquiry into Google’s AI model to ensure compliance with EU regulations. The focus is on Google’s Data Protection Impact Assessment (DPIA) and its processing of personal data under the General Data Protection Regulation (GDPR). This reflects broader efforts by EU and … Read more
September 12, 2024 at 07:18AM The Irish Data Protection Commission has initiated a statutory inquiry into Google’s AI model, PaLM 2, to assess its compliance with data protection regulations when processing European users’ personal data. This inquiry aims to safeguard individuals’ rights and freedoms, following similar actions against other tech companies for potential privacy violations. … Read more
June 28, 2024 at 02:47AM Microsoft published details about the Skeleton Key technique, which bypasses safety mechanisms in AI models to generate harmful content. This could prompt AI models to provide instructions for creating a Molotov cocktail. The technique highlights the ongoing challenge of suppressing harmful content within AI training data, despite efforts by companies … Read more
June 20, 2024 at 08:32AM AI model weights control system outputs, but if altered or “poisoned,” they can produce erroneous and potentially dangerous results. Securing these weights is crucial for maintaining the reliability and safety of artificial intelligence systems. This was highlighted in a post on SecurityWeek. Based on the meeting notes, the key takeaway … Read more
June 14, 2024 at 01:06AM Microsoft has delayed the rollout of the AI-powered Recall feature for Copilot+ PCs, shifting to use the Windows Insider community to ensure quality and security. Originally set for a broad release on June 18, it faced controversy over privacy and security concerns. The move underscores the careful balance needed for … Read more
June 7, 2024 at 12:40PM Microsoft is updating its AI-powered Recall feature for Copilot+ PCs following customer pushback. The feature will now require user opt-in and Windows Hello verification for use. Additionally, Recall’s data encryption and security measures are being enhanced. These updates will be released with Recall’s preview on June 18, with potential changes … Read more
May 28, 2024 at 11:12AM OpenAI announced the establishment of a safety and security committee to advise on critical decisions for its projects and operations. This comes amidst debate on AI safety, following resignations and criticism from researchers. The company is training a new AI model and claims industry-leading capability and safety. The committee, including … Read more
May 21, 2024 at 07:09AM A critical security flaw in the llama_cpp_python Python package (CVE-2024-34359, codenamed Llama Drama) allows threat actors to execute arbitrary code, posing a risk to data and operations. Another high-severity flaw in Mozilla’s PDF.js library permits JavaScript execution in the PDF.js context. Both issues have been addressed in recent software updates. … Read more