The lack of rate-limiting and duplicate request filtering in the API exacerbates the risk of DDoS attacks, as it allows for massive simultaneous requests to targeted sites.

When multiple hyperlinks are submitted, OpenAI's servers send out an HTTP request for each link, resulting in a surge of requests that can overwhelm the target site.

The potential for successful DDoS attacks raises alarms about website downtime, loss of revenue, and damage to company reputations, underscoring the urgency for OpenAI to address the issue.

This incident highlights the necessity for improved communication and transparency between cybersecurity researchers and technology firms, as well as the need for better security practices.

Experts recommend that companies implement regular security audits, code reviews, and penetration testing to identify and rectify potential weaknesses in their systems.

Elad Schulman, CEO of Lasso Security Inc., supports Flesch's findings, warning that such vulnerabilities can lead to reputational damage and financial exploitation.

A recent vulnerability in OpenAI's ChatGPT API could potentially enable distributed denial of service (DDoS) attacks, as highlighted by cybersecurity researcher Benjamin Flesch.

The issue arises from the API's handling of HTTP POST requests at the /attributions endpoint, where it allows an array of hyperlinks without limiting their number or verifying duplicates.

Discovered in early January 2025, this vulnerability was reported to both OpenAI and Microsoft, but as of January 10, there has been no acknowledgment or effective response from either company.

Flesch attempted to notify both companies multiple times through various channels, including OpenAI's BugCrowd platform, but claims he received no acknowledgment or updates.

Flesch rated the severity of this vulnerability at 8.6 on the Common Vulnerability Scoring System (CVSS), indicating it is easy to exploit and poses a significant risk to service availability.

In addition to the DDoS vulnerability, Flesch noted that the API is also susceptible to prompt injection attacks, which could allow malicious users to manipulate its responses.