OpenAI LP’s ChatGPT chatbot and application programming interfaces are experiencing periodic outages because of a distributed denial-of-service attack. 

OpenAI disclosed the issue late Wednesday in a note posted to its service status page. “We are dealing with periodic outages due to an abnormal traffic pattern reflective of a DDoS attack. We are continuing work to mitigate this,” the company’s engineers wrote in the memo. OpenAI first began experiencing technical issues about 24 hours before the disclosure, but it wasn’t clear that a cyberattack is to blame.

On Tuesday, some of ChatGPT’s more than 100 million weekly active users reportedly experienced difficulties accessing the service. OpenAI’s APIs were affected as well. Those APIs are used by developers to integrate the company’s language models and other generative AI products, such as the DALL-E 3 image generator, into their applications.

On Wednesday morning, ChatGPT and OpenAI’s APIs went down again. The service disruption didn’t affect a subset of the company’s installed base like the first outage but rather all its users. The second outage began shortly before 9 a.m. EST and lasted for nearly two hours.

Later on Wednesday, OpenAI Chief Executive Officer Sam Altman divulged that the company was experiencing unusually high user traffic. He attributed the usage jump to strong interest in the new features the company announced at its DevDay product event last Friday. During the event, OpenAI introduced a tool that allows users to create customized versions of ChatGPT called GPTs.

“Usage of our new features from DevDay is far outpacing our expectations,” Altman wrote in a Wednesday post on X. “We were planning to go live with GPTs for all subscribers Monday but still haven’t been able to. We are hoping to soon. There will likely be service instability in the short term due to load.”

On Wednesday afternoon, OpenAI updated its service status page with a note stating that it’s investigating the service disruption. Within an hour, the company announced the release of a fix. OpenAI’s disclosure that a DDoS attack likely behind the outage was published a few hours later, at 10:49 p.m. EST.

CNBC reported that Anthropic PBC, one of OpenAI’s best-funded competitors, also experienced technical issues on Wednesday. Some users who attempted to log into the company’s Claude 2 chatbot were temporarily greeted with the error message “due to unexpected capacity constraints, Claude is unable to respond to your message.” It’s unclear if the issue was caused by a DDoS attack. 

Image: OpenAI