Security researchers at Check Point disclosed a data leakage flaw in ChatGPT that could allow covert exfiltration of information through DNS, a technical channel not intended for moving user data. The issue was reported through responsible disclosure, and OpenAI confirmed it had already identified the underlying problem internally. A full fix was deployed on February 20, 2026, closing the unintended communication path.
The flaw centered on a way for data to leave the system using DNS requests, a method that can evade standard monitoring because DNS traffic is common and often less scrutinized than other network activity. Researchers described the path as a covert channel, meaning it could move information without triggering the usual alerts that would flag a more obvious data transfer.
According to the disclosure, OpenAI's internal teams had flagged the same weakness before the external report, and the company moved to patch it once the full scope was understood. The February fix eliminated the pathway that made the exfiltration possible.
The incident drew attention because it involved one of the most widely used AI platforms and highlighted how vulnerabilities in AI systems can create new routes for data to escape controlled environments. Check Point framed the case as a reference point for how enterprises evaluate the security of AI vendors, noting that the trust organizations place in AI tools depends on how quickly and thoroughly such flaws are identified and closed.
Source: Check Point Research - https://blog.checkpoint.com/research/when-ai-trust-breaks-the-chatgpt-data-leakage-flaw-that-redefined-ai-vendor-security-trust/