It seems that OpenAI is struggling to keep its ChatGPT service online, given how frequently the GPT-3.5 and GPT-4 powered chatbot has been down this past couple of weeks.
Over the few past day, users of this widely used AI chatbot have been confronted with frustrating outages. These disruptions have manifested as login difficulties and an unwelcome message, declaring, “ChatGPT is at capacity right now.” Or, in some cases, it will simply say, “Something went wrong…”
Turns out, OpenAI may have been facing persistent DDoS attacks. In their latest update, issued late Wednesday evening, OpenAI attributed the “periodic outages” to a potential cyberattack, suggesting that these disruptions were provoked by an “abnormal traffic pattern reflective of a DDoS attack.”
For those who are not aware, a DDoS attack or distributed denial-of-service attack occurs when hackers overwhelm a server with a deluge of traffic.
The exact cause of these outages, or the source of the attacks if there is indeed one, remains uncertain. However, tech news site TechCrunch, based on Telegram’s chats they received, has reported that a hacktivist group called Anonymous Sudan, associated with the pro-Russian hacktivist collective Killnet, claimed responsibility for the disruptions.
Ultimately, this purported cyberattack underscores the enduring infrastructural vulnerabilities that can affect even the most prominent and technologically advanced internet entities. It serves as a reminder that maintaining the daily operations of a website, even one housing groundbreaking AI technology, can pose significant challenges.
As to what caused the attack, a member of the hacktivist group claimed on Telegram that they targeted ChatGPT due to perceived bias against Palestine and in favour of Israel, as per a report by Forbes.
OpenAI CEO Sam Altman initially attributed the outages to increased user interest following the company’s announcement on Monday of new features, including GPT-4 Turbo, their most powerful AI model to date, and tools for users to create personalized “GPT” AI models. Altman tweeted on Wednesday afternoon that the usage of these new features was surpassing expectations, and that “there will likely be service instability in the short term due to load.”
Notably, the issues on Wednesday did not exclusively affect OpenAI’s ChatGPT. According to CNBC, Claude 2, a chatbot operated by OpenAI’s competitor Anthropic, also encountered problems. If not anything else, the timing of the attack on these two services seems to indicate that this was indeed a concerted and orchestrated attack.
While the potential cyberattack and its ramifications persist, OpenAI CEO Sam Altman has yet to comment. Many users and stakeholders are urging the company to provide more transparency, considering the widespread popularity of the ChatGPT service