While tech giants continue to champion the safety of their AI systems, a startling discovery has revealed that OpenAI’s ChatGPT isn’t as secure as many believed. A devastating server-side request forgery vulnerability, known as CVE-2024-27564, has been exploited in over 10,479 attack attempts in just one week. All from a single malicious IP address. Let that sink in.
The vulnerability affects ChatGPT’s pictureproxy.php component, allowing attackers to inject malicious URLs into parameters and force unintended requests. With a CVSS score of 6.5, it’s classified as “medium” severity. Medium. Right. Tell that to the victims.
A medium-severity vulnerability that’s exploited 10,000+ times in a week isn’t “medium” to those under attack.
US government organizations bore the brunt of these attacks, accounting for 33% of all exploitation attempts. Germany and Thailand weren’t far behind, each representing 7% of targets. Indonesia, Colombia, and the United Kingdom also made the hit list.
The cybersecurity firm Veriti documented a surge in January 2025, followed by a slight decrease in February and March.
What’s truly alarming? About 35% of analyzed companies remain vulnerable due to misconfigurations in their security systems. IPS, WAF, firewall – basic protections botched. Financial institutions are particularly at risk, with potential consequences including data breaches, unauthorized transactions, and regulatory penalties.
The proof-of-concept exploit code is publicly available. Anyone can use it. No authentication required. Just point and shoot.
This incident highlights how determined attackers can weaponize even medium-severity vulnerabilities. The exploitation was first reported in September 2023 and publicly disclosed a year ago, yet organizations continue to fall victim.
The broader implications are clear: as AI systems and API integrations become more prevalent, our attack surface expands. A proper risk assessment could have identified these vulnerabilities before they were exploited on such a massive scale. Medium-severity vulnerabilities are too often overlooked. But attackers? They don’t overlook anything. Banks and fintech companies face heightened risk due to their reliance on AI services and extensive API integrations. Thankfully, there has been no evidence of a breach in OpenAI’s infrastructure despite these attacks.
For victims, the consequences extend beyond technical issues to reputational damage and financial losses. It’s a stark reminder that in cybersecurity, there’s no such thing as “just a medium vulnerability.” Not when there’s 10,000+ attacks in a week.
