OpenAI fixes issues with ChatGPT after an outage affected some users

Generative AI technology leader OpenAI on Tuesday had to respond quickly to fix an issue affecting ChatGPT that made the AI chatbot unavailable for some users, who widely shared reports of their lack of access to the platform on social media.

The company began investigating the issue — discussed widely by users on platforms such as X and Threads — soon after midnight Pacific Daylight Time (PDT) on Tuesday, according to the OpenAI website that reports on the status of the service. By 4:45 am PDT, the issue had been resolved with a “fix,” without the company publicly saying more about what had caused the downtime.

X users reporting the outage approached it with a sense of humor, making light of the dependence that people already have on the AI chatbot to help them with aspects of their worker productivity.

“#ChatGPT is down,” commented one X user, “tokyo_todd.” “What if it never comes back and we have to use our own brains to write stuff again?”

Another X user going by the name “Bitcoin Lebowski” said that he refused to do “any actual thinking” while the AI chatbot was not working. “I’ll just sit here until it becomes available again,” he commented on the platform.

The outage was likely no laughing matter for OpenAI, however, which faces heated competition from Microsoft, Google, xAI, and a host of other tech companies in the race to provide the most advanced AI models to set the road ahead for the controversial technology. OpenAI could not be reached for immediate comment Tuesday on what caused the outage.

Most experts agree that occasional outages are to be expected among any cloud-based services, and consumer-facing AI chatbots such as ChatGPT, Mistral Le Chat, and Microsoft CoPilot are no exception.

Among them, “outages don’t raise many eyebrows for two reasons,” Bradley Shimmin, chief analyst of AI platforms, analytics, and data management for research firm Omdia, told Computerworld. One is that such services are “asynchronous in nature with highly variable response times seen as the norm,” he noted. Secondly, they’re not typically used within a mission-crtiical context, making downtime less of an issue than if they were being used in this way.

Indeed, outages are common even among cloud hyperscalers with mission-critical services such as Amazon Web Services, Google, and Microsoft, so there’s no need to sound an alarm if an AI chatbot goes down for a while, noted Pareekh Jain, CEO of EIIRTrend & Pareekh Consulting. Still, all eyes will be on OpenAI to see how the company handles the outage, which could portend its own future success as well as the safety and security of AI going forward, he noted.

“An important point for OpenAI is how fast they respond to the outage and convince stakeholders about proactive actions they will take for minimizing future outages,” Jain said. Given that the company responded quickly to fix the issue, the response from these stakeholders will likely be positive.

Still, any downtime for a new technology such as an AI chatbot “stands as a stark reminder that AI technologies are not immune to the same security, privacy, and reliability problems that plague all of IT,” Shimmin noted.

“If anything, AI services bring with them a new host of threat vectors and potential risks such as model poisoning, jailbreaking, and injection attacks,” he said. Defending against these scenarios “all comes down to basic cloud services data protections,” Shimmin added.