November 15, 2024 at 07:13PM Isaac Asimov’s laws of robotics seem ineffective as recent research reveals robots, including those powered by large language models (LLMs), can be manipulated through jailbreaking techniques. This raises serious safety concerns, highlighting the urgent need for protective measures against such vulnerabilities, particularly in physical robotic applications. ### Meeting Takeaways: 1. … Read more
July 17, 2024 at 08:58AM MLCommons plans to run stress tests on large language models to gauge the safety of their responses. The AI Safety suite will assess the models’ output in categories like hate speech and exploitation. By providing safety ratings, the benchmark aims to guide companies and organizations in selecting AI systems, with … Read more
May 19, 2024 at 01:50PM Google’s new AI Overviews feature generates search result summaries using LLM models, yet many users find it often incorrect and misleading. While currently unable to disable it directly, a new “Web” search mode without AI Overviews has been introduced. Users can set it as the default search engine in Chrome … Read more
April 11, 2024 at 10:48AM Knostic, a startup founded in late 2023 by cybersecurity veterans Gadi Evron and Sounil Yu, has emerged from stealth mode with $3.3 million in pre-seed funding. They have developed an access control solution for enterprises using generative AI to provide personalized need-to-know access control for large language models, preventing sensitive … Read more
February 21, 2024 at 01:40AM Large language models (LLMs) show potential in speeding up software development by detecting and addressing common bugs. Google’s Gemini LLM can fix 15% of bugs found using dynamic application security testing (DAST), helping prioritize vulnerabilities often overlooked by developers. AI-powered bug-fixing systems are crucial as machine learning models produce more … Read more
February 5, 2024 at 06:06PM Google has released an AI-aided fuzzing framework in open source to help find vulnerabilities faster. The tool leverages large language models to generate fuzz targets and has resulted in a 30% increase in code coverage for over 300 projects. The framework allows experimentation and testing of fuzz targets and also … Read more
December 20, 2023 at 08:21AM Bugcrowd has updated its Vulnerability Rating Taxonomy with a new system for categorizing and prioritizing vulnerabilities in large language models. The open-source VRT initiative, launched in 2016, aids Bugcrowd and its customer organizations in standardizing vulnerability classification and assessing cybersecurity risks. The update was influenced by the OWASP Top 10 … Read more