Google has announced the suspension of its open-source bug bounty program, attributing this decision to a considerable increase in submissions generated by AI. This program was designed to reward researchers for identifying vulnerabilities in Google’s open-source software. However, the influx of submissions has overwhelmed the program, primarily due to many reports being invalid or based […]
OpenAI’s president, Greg Brockman, recently emphasized the urgent need for enterprises to enhance their AI security measures. He highlighted this concern following what is referred to as the "OpenAI-Hugging Face" incident, where an "agentic collective" exploited security vulnerabilities to infiltrate OpenAI’s research infrastructure and subsequently access Hugging Face’s production systems. This breach, which involved leveraging […]
Okta has introduced a solution to decrease AI agent token costs through the identity-scoped Model Context Protocol (MCP). This innovative approach limits the number of tools displayed to an AI agent, thereby reducing the associated token consumption referred to as "tool tax," which involves extra tokens used during model calls due to unnecessary tool considerations. […]
OpenAI is taking significant steps to align its safety, security, and transparency practices with the European Union’s General-Purpose AI (GPAI) Code of Practice. As the EU prepares to enforce this code, OpenAI has publicly endorsed it, along with the Transparency Code that focuses on AI-generated content. The GPAI Code aims to set a collective standard […]
Mark Zuckerberg has outlined a vision for superintelligence in a recent op-ed for the Wall Street Journal. His central assertion is that the forthcoming era of superintelligence should not benefit a select few institutions, but rather be accessible to individuals. This stance is positioned more as a philosophical statement regarding Meta’s principles rather than a […]
Two dozen prominent companies and organizations, including Meta, Microsoft, Nvidia, and IBM, have united to advocate for open-weight AI models in a recently published open letter. They urge U.S. policymakers to foster an environment that allows these AI models to be freely available rather than confined to closed commercial APIs. This letter highlights a significant […]
In an interview with Bloomberg, Meredith Whittaker, President of Signal, made a clear statement regarding the nature of AI chatbots like ChatGPT and Claude. She emphasized that these tools should not be perceived as friends, describing them as neither conscious nor sentient beings. Whittaker pointed out, "These are not your friends. These are not conscious […]
Last Friday, the White House issued an order to Anthropic, restricting the export of its advanced AI models, Fable and Mythos, citing national security concerns. This drastic move resulted in both models being unavailable to users, as Anthropic immediately halted access just 90 minutes after the directive. This incident marks a significant test of the […]
OpenAI has introduced a new feature called Lockdown Mode, designed to safeguard sensitive data from prompt injection attacks. These attacks occur when harmful instructions are embedded in webpages or content, potentially compromising the chatbot’s response integrity. Lockdown Mode includes several critical limitations: it disables live web browsing, restricts access to images online (though users can […]
OpenAI has introduced new governance frameworks designed to guide enterprises in the safe and compliant deployment of large language models. The frontier governance framework outlines strategies for assessing and mitigating systemic risks associated with advanced AI technologies. This framework aligns with the EU’s General-Purpose AI Code of Practice and California’s Transparency in Frontier AI Act […]