AI start-up Anthropic says Claude is back up and running after hundreds of users reported the AI chatbot and coding assistant experienced issues for roughly two hours on Friday afternoon.

Anthropic said it was also investigating social media claims that Claude may have returned responses meant for other users during the disruption, raising concerns about a possible customer data leak.

The disruption hit several of Anthropic’s frontier Opus AI models starting at about 11:19 a.m. EDT (15:19 UTC), prompting an immediate investigation.

Anthropic confirmed elevated errors affecting Claude API, Claude Code, Claude AI, and Claude Cowork services.

The service interruption caused “elevated errors on many Claude models,” impacting chatbot performance and response success rates, Anthropic said in an update on its Claude status page.

Engineers identified the issue roughly 20 minutes later.

Multiple Claude AI models affected

Anthropic listed the affected products alongside the time each issue was discovered and later resolved. The company said five frontier models in total were affected, including its flagship Opus large language model and the more advanced multimodal Sonnet.

• Opus 4.6 — 8:25 PT / 15:25 UTC
• Sonnet 4.6 — 9:23 PT / 16:23 UTC
• Opus 4.8 — 9:59 PT / 16:59 UTC
• Opus 4.7 — 10:12 PT / 17:12 UTC
• Opus 4.5 — 10:29 PT / 17:29 UTC

The San Francisco-based AI company said all model issues were resolved by 2:27 p.m. EDT (18:27 UTC).

“Success rates across all models have returned to expected levels. We are continuing to monitor closely to ensure no further issues will recur,” Anthropic said.

Users began reporting problems with the service on the internet-outage monitoring site Downdetector, which showed nearly 1,000 reports in the US at its peak.

According to the site, 40% of users had problems with Claude Chat, 33% had issues with Claude Code, and another 20% reported interruptions with the Claude app.

As is often the case during AI outages, some developers voiced frustration on social media over repeated Claude disruptions.

“Another Friday, another day where both Claude Code and the Claude web app are down for me,” Yuchen Jin, a former AI start-up CTO, posted on X.

“I’m starting to understand why Mythos still hasn’t shipped. Hello Codex,” they said.

Anthropic investigating customer data leak claims

Meanwhile, at about 12:00 p.m. EDT, one AI analyst claimed on X that Claude may have “returned another user’s inference output” during the outage, raising concerns about a possible customer data exposure incident.

The original post, made by New York City-based software engineer Moritz Wallawitsch, had racked up over a million views by Friday evening.

After contacting Anthropic for clarification, a spokesperson told Cybernews the company had "no other reports of customer data leaking” but was now investigating the claim.

The outage was “completely unrelated and would not have impacted customer data,” the spokesperson said, referring Cybernews to comments made by Claude Code engineer Thariq Shihipar.

“This incident earlier today was related to an infrastructure outage, not a bug with our inference,” Shihipar wrote in response to the X post, just after 6:00 p.m. EDT.

“We haven't seen any other reports or evidence of customer data being leaked, but take any reports of this very seriously and are investigating,” Shihipar said, adding that he was sending Wallawitsch a direct message about the matter.

This incident earlier today was related to an infrastructure outage, not a bug with our inference.

We haven't seen any other reports or evidence of customer data being leaked but take any reports of this very seriously and are investigating.

I sent you a DM to get more info.

undefined Thariq (@trq212) June 5, 2026

Still, Claude users replying to the thread appeared split over whether the chatbot was returning answers meant for other users or simply producing nonsensical responses.

Some users said they were not surprised, suggesting cross-user leakage on the AI platform was not isolated to this incident. Others said it was more likely Claude was simply hallucinating.

Recent Claude disruptions raise reliability concerns

Friday’s disruption follows several other elevated-error incidents affecting Anthropic’s Opus models over the past few months, including several in the past week alone.

Earlier this week, on June 3rd, Opus 4.7 experienced a brief disruption due to elevated errors, and more significantly, some Claude Code services were degraded for more than 3 hours overnight.

Anthropic said the degraded services had impacted Claude Code security reviews, code reviews, routines, and some Claude Code web sessions.

Errors also triggered disruptions across multiple models for about 5 hours and 45 minutes on June 2nd, while several Claude Sonnet and Opus models suffered elevated request errors. Similar degraded services were also reported on June 1st, according to Anthropic’s status page.

In March, Anthropic’s Claude suffered a massive worldwide outage, impacting thousands of users across the US, the UK, Australia, and parts of Europe. The company blamed the disruption – since resolved – on the failures of “some API methods.”

And more recently, on April 7th, Anthropic scrambled to fix Sonnet 4.6 and Claude AI models after multiple outages spanned a two-day period. The reason behind those outages was never revealed.

Unlock more exclusive Cybernews content on YouTube.