Security Flaw in ChatGPT Share Links Lets Attackers Hijack Sessions and Access Private Conversations

Security researchers at the firm Parameter have uncovered a vulnerability in OpenAI's ChatGPT share link feature that could let malicious actors execute code within a logged‑in user’s browser session and harvest the contents of private chats.

The flaw stems from the way shared links embed executable scripts. When a user opens a crafted link while signed into ChatGPT, the malicious payload runs in the same context as the legitimate site, effectively giving the attacker the same privileges as the user. This includes the ability to read conversation history, extract API tokens, and potentially manipulate subsequent interactions.

Parameter reported the issue to OpenAI through the company’s vulnerability disclosure program. Although OpenAI has not publicly detailed its remediation steps, the firm typically patches such bugs promptly and may roll out additional safeguards to isolate shared content from active sessions. The discovery was first highlighted by The Hacker News, which prompted a broader discussion among security analysts about the risks of convenience features that blur the line between user‑generated content and executable code.

ChatGPT’s share link function was introduced to make it easier for users to showcase AI responses without requiring recipients to have an account. While the feature has been praised for its utility in education and collaboration, it also opens a surface for abuse if not properly sandboxed. Similar cross‑site scripting (XSS) concerns have surfaced in other web platforms that allow embedding of dynamic content, underscoring the need for rigorous input sanitization.

Experts advise users to treat shared links with the same caution applied to any external URL. Logging out of ChatGPT before clicking unknown links, using private browsing modes, or employing browser extensions that block third‑party scripts can reduce exposure. Organizations that rely on ChatGPT for internal communications may also consider policy controls that restrict the use of share links on corporate devices.

As AI tools become more integrated into daily workflows, the incident highlights a growing tension between usability and security. Ongoing monitoring, prompt disclosure of vulnerabilities, and transparent communication from providers like OpenAI are essential to maintaining user trust while the technology continues to evolve.