OpenAI fends off massive DDoS attack to finally launch GPTs

OpenAI’s post-DevDay victory lap just hit an unexpected wall of malicious traffic.

On Thursday, CEO Sam Altman announced that GPTs, the company's new customizable AI chatbots, are finally live for all paying subscribers. The rollout came right after engineers fended off a sustained DDoS attack that knocked ChatGPT and its API offline worldwide.

It was not just high demand taking down servers

Altman had originally planned to launch GPTs for all subscribers on Monday. But when widespread outages crippled ChatGPT and developer tools on Wednesday, the timeline collapsed.

At first, Altman blamed the chaos on sheer popularity, claiming that post-DevDay traffic for features like GPT-4 Turbo and cheaper pricing had far outpaced expectations. That turned out to be only part of the story.

By Wednesday evening, OpenAI’s outage tracker admitted the company was dealing with abnormal traffic patterns consistent with a coordinated DDoS attack. Attackers flooded OpenAI's servers with traffic from infected device networks, making it impossible to separate legitimate queries from malicious pings.

The hacker collective Anonymous Sudan quickly claimed responsibility on Telegram, as first spotted by Bleeping Computer. Cybersecurity firm Truesec previously linked the group to Russian-backed operations, noting its history of attacking Microsoft and Swedish institutions.

While Anonymous Sudan's claim remains unconfirmed, the group stated the attack was a demonstration of support for Palestine, specifically targeting OpenAI over ties to Israel following Altman's trip to Tel Aviv in June, where he noted Israel’s massive future role in AI.

Why it matters

For regular users and builders, GPTs are officially here, meaning anyone can start building custom chatbots without touching code. But this multi-day disruption is a harsh reminder that building critical workflows on a single proprietary platform comes with serious downtime risks right now.

For the broader AI ecosystem, this incident marks a turning point. Frontier AI labs are no longer just software darlings; they are now prime geopolitical and ideological targets operating high-value infrastructure.

We think this mess exposes a real gap between OpenAI’s product ambitions and its infrastructure resilience. Pitching an App Store-style developer ecosystem while getting your front door jammed by standard DDoS tactics is not a great look. If OpenAI wants to be the world's backbone for AI software, it looks like it will have to invest just as heavily in basic enterprise-grade defense as it does in next-generation model training.