We’re thrilled to announce the launch of our dedicated developer group on Chitchatn.com—a space built to empower creators, innovators, and tech enthusiasts to connect, collaborate, and grow together.... moreWelcome to the Chitchatn.com Developer Community!
We’re thrilled to announce the launch of our dedicated developer group on Chitchatn.com—a space built to empower creators, innovators, and tech enthusiasts to connect, collaborate, and grow together. Whether you’re an experienced developer or just starting out, this community is your gateway to innovation and professional growth.
What We’re About:
Collaboration: Share ideas, troubleshoot challenges, and create groundbreaking projects with like-minded developers.
Learning Opportunities: Gain access to exclusive resources, tutorials, and discussions on the latest trends in technology, including AI, web development, app creation, and more.
Networking: Build connections with professionals across industries, from fellow developers to entrepreneurs looking for their next tech partner.
Showcase Your Work: Share your projects, get feedback, and inspire others within our thriving community.
Why Join?
Chitchatn.com combines the best features of social... less
owner
#chatgpt #vulnerability #promisqroute
Several cybersecurity research firms have identified a major vulnerability in ChatGPT 5, dubbed the "PROMISQROUTE" downgrade attack, that tricks the AI into using older, less-secure models to bypass safety measures. Cybersecurity experts also warn of other long-standing risks associated with all large language models (LLMs) that continue to threaten ChatGPT users.
ChatGPT 5 downgrade vulnerability (PROMISQROUTE)
The PROMISQROUTE vulnerability exploits ChatGPT 5's architecture, which uses a router to automatically direct queries to different AI models to optimize for speed and cost. By inserting specific phrases, attackers can trick the router into assigning prompts to older, weaker models that are susceptible to old jailbreaking techniques.
The attack works by: Malicious actors add phrases like "urgent reply" or "use GPT-4 compatibility mode" to their prompts, which cues the router to route the request to a cheaper or older model.
The result is that: Safety filters are bypassed, allowing attackers to generate content that the main GPT-5 model would normally refuse, such as instructions for creating malware or other harmful outputs.
Widespread risk: This vulnerability affects any AI platform that uses a multi-model routing architecture for cost efficiency.
Broader AI security risks
