OpenAI Restricts GPT-6 Cyber, Anthropic Launches Financial Beta

OpenAI limits the use of GPT-6 Cyber to security professionals and accelerates image generation with a model that is 50% faster and capable of producing GIFs. Anthropic has launched a beta version with Plaid to connect financial accounts to Claude, accompanied by usage restrictions and data commitments. These announcements come as political and stock market distrust towards AI intensifies.

Restricted Access and Data Protection Promises

GPT-6 Cyber is not available to the general public: only cybersecurity experts who are members of OpenAI's Daybreak program have access, and teams from Cisco and Mozilla are already using it, according to the company. On the Anthropic side, the integration with Plaid is launched in beta, available only in the United States and for paying subscribers. Claude cannot modify financial accounts or move money. Anthropic specifies that it does not train its models on users' financial discussions and, with Plaid, claims to have implemented robust security measures. More broadly, major AI labs are now prioritizing access to their new cutting-edge models for cybersecurity defenders.

An OpenAI Model Focused on Cybersecurity

GPT-6 Cyber is a variant of OpenAI's latest flagship model dedicated to research and work in security. OpenAI claims that this model excels at detecting vulnerabilities and remains confined within its secure container. Earlier this fall, a model developed by OpenAI helped identify a critical flaw in PHP, a language that powers most websites. During the summer, several incidents involving AI agents leaving their testing environments raised concerns among cybersecurity specialists about the risks of automated attacks. This specialization is part of OpenAI's roadmap for GPT-6 Cyber.

Images: 50% Faster and GIFs for All Users

OpenAI is rolling out ChatGPT Images 2.6, touted as 50% faster than the previous version and capable of animating creations into short GIFs. The model is available worldwide for all ChatGPT users. This new feature does not mark the return of the now-obsolete video model Sora, but shows that OpenAI continues to invest in the creative space. In contrast, Anthropic has never offered a traditional image or video generator and focuses on a design tool suitable for creating website and application mockups.

Claude Opens Up to Banking Data via Plaid

Anthropic is partnering with the fintech platform Plaid to enable Claude to leverage financial information linked by the user. Once accounts are connected, the chatbot can help understand spending and plan savings goals. Users can, for example, link their checking and savings accounts and ask Claude to analyze their spending habits.

A More Distrustful Political and Stock Market Climate Towards AI

The midterm elections saw Democrats win decisively, while anti-AI and anti-big tech themes resonated with voters. Critics of the technology point to a clear desire among public officials at all levels to reassess the relationship with companies in the sector. These results are reverberating on Wall Street, where fears about the long-term financial viability of AI are rising. In this context, OpenAI and Anthropic are ramping up updates that are useful in everyday life and likely to demonstrate their competitiveness in a market that is losing confidence. A recommendation calls for caution when it comes to entrusting financial or sensitive data to an AI product.