AI Pulse · Frontier Labs Watch

Claude had another platform-wide outage at 05:06 UTC today, affecting claude.ai, the API, Claude Code, and Cowork at the same time. The affected services reportedly included Mythos 5, Fable 5, Opus 5, and Opus 4.8. When that many products and models fail together, it points less to a problem with one model and more to a shared layer, such as routing or core platform infrastructure.

The bigger concern is the pattern. This incident is being counted as the 28th disruption in 30 days for Claude’s commercial tier, while Claude for Government is reported to have maintained 100% uptime over the past 90 days. If those figures hold, business customers are seeing a very different reliability standard from government customers. That matters if Claude sits inside customer support, reporting workflows, software development, or internal operations. An outage is frustrating. Repeated outages force teams to build workarounds, hold backup tools, and question whether a platform can handle a critical process.

For owners and operators, the lesson isn’t to abandon any one AI provider. It’s to avoid putting a single provider at the centre of an important workflow without a fallback. Keep a second model available, define what staff do when the primary tool is down, and monitor failures by service, not just by vendor. This is the kind of thing we build into an AI command centre, so teams can see where a dependency has failed and switch work accordingly.