Anthropic faces Claude failures and suspends Mythos 5 and Fable 5 models

On June 23, 2026, Anthropic began investigating an increase in errors in several Claude models at 14:19 UTC, according to its status page. The company reported that it has identified a fix for the failures, although full service has not yet been restored. At the same time, the company is still explaining why it suspended access to Claude Mythos 5 and Claude Fable 5, two models that had been recently launched with great anticipation. This double event — a widespread outage and the suspension of two flagship models — has generated a storm of uncertainty in the enterprise AI ecosystem.

What happened?

According to Anthropic's status page, the investigation began at 14:19 UTC on June 23, when “elevated error rates” were detected in multiple Claude models. At 16:45 UTC, the company updated the status to “identified,” stating that it had found the root cause and was implementing a fix. However, no timeline for full service restoration has been provided. This incident comes just two weeks after Anthropic suspended access to Claude Mythos 5 and Claude Fable 5, launched in April 2026 with great fanfare. The company has been vague about the reasons, only saying that “further evaluation is needed” before resuming access. Internal sources cited by The Next Web suggest the models may have shown unpredictable behaviors in complex reasoning tasks, leading to a safety review.

Why is this important?

This incident comes at a delicate time for Anthropic. The suspension of Mythos 5 and Fable 5 had already raised doubts among enterprise users, who rely on Claude's stability for critical tasks such as contract analysis, financial report generation, and automated customer support. Now, the widespread failures worsen the perception of unreliability. In a market dominated by OpenAI and Google, any disruption can translate into loss of customers and credibility. According to a May 2026 Gartner report, 78% of companies adopting generative AI prioritize availability and consistency over model capability. Anthropic, which had gained traction with Claude 3.5 Sonnet and Claude Opus 4, now sees its position in the enterprise segment threatened, where it directly competes with GPT-4o and Gemini Ultra.

Consequences for the market and users

For companies integrating Claude into their workflows, these errors can cause delays in automated processes, loss of productivity, and additional costs. For example, a bank using Claude for fraud detection could see its operations disrupted, while an ecommerce platform using Claude for customer support chatbots could experience increased response times. Developers building on Anthropic's API see their service continuity compromised, and some have already reported on forums like Hacker News that they are temporarily migrating to alternatives such as OpenAI's API or Mistral's open-source models. In the long term, trust in the platform could erode if these incidents recur. An IDC analyst estimates that each hour of downtime for an enterprise AI API can cost client companies between $10,000 and $50,000 in lost productivity, not to mention reputational damage.

What should readers know?

• Anthropic has identified a fix for the elevated errors but has not given a restoration timeline. The status page updates every 30 minutes, but users should be prepared for a prolonged outage.
• The suspension of Mythos 5 and Fable 5 remains without detailed explanation, suggesting possible safety or performance issues. Anthropic has said it will “share more details in the coming weeks,” leaving an information vacuum that fuels speculation.
• Users should monitor Anthropic's status page and consider contingency plans, such as having backup models or agreements with multiple AI providers.
• This event underscores the importance of redundancy in critical AI tools. Companies should evaluate diversifying their AI providers to mitigate the risk of disruptions.

“Trust is the most valuable asset in enterprise AI. Incidents like this can have a disproportionate impact on adoption.” — Analyst at TheVortiq

Historical context

This is not the first time Anthropic has faced availability issues. In January 2026, a massive Claude outage affected thousands of users for several hours due to a storage system failure. On that occasion, the company restored service within 4 hours and published a detailed post-mortem report. However, the simultaneous suspension of two flagship models — Mythos 5 and Fable 5 — is unprecedented. These models, launched in April 2026, were acclaimed for their performance on reasoning and creativity benchmarks, surpassing GPT-4o in several tests. Their sudden suspension, without a clear explanation, has led to speculation about potential security vulnerabilities, unmitigated biases, or even alignment issues. The company will need to communicate the causes transparently to prevent the perception that Anthropic has serious internal quality control problems from taking hold.

Comparisons with competitors

OpenAI has had similar outages, such as the ChatGPT downtime in November 2025 that lasted 3 hours, but its recovery and communication track record is more robust. OpenAI typically publishes incident reports within 24 hours and offers compensation to enterprise clients in the form of credits. Google, for its part, maintains near-perfect availability for Gemini, with a 99.95% SLA for its enterprise API, backed by its cloud infrastructure. Anthropic, which uses a mix of its own data centers and AWS cloud services, has yet to publish a public SLA for Claude, creating uncertainty among customers. To compete in the enterprise segment, Anthropic needs to demonstrate that it can match those standards of reliability and transparency. Otherwise, it risks being relegated to a niche of enthusiast users, while large enterprises opt for more established providers.