Anthropic just auto-deployed a more autonomous AI to millions of users. Your change control process just failed silently.

On June 30, Anthropic made Claude Sonnet 5 the default model for every Free and Pro user.

No announcement to enterprise admins. No change control gate. No deployment readiness review.

Your organization just got a substantially more autonomous AI system because a vendor flipped a switch.

Sonnet 5 brings browser control, terminal access, and multi-step autonomous task execution to default tiers.

These are capabilities that required Opus-class models three months ago.

The AI Governance Institute put it bluntly: organizations that have not explicitly pinned model versions are now running a more autonomous system "without any deliberate deployment decision having been made."

This triggers obligations under AI model change management and pre-production approval policies that most enterprises haven't built yet.

63% of organizations cannot enforce purpose limitations on AI agents. 60% cannot terminate a misbehaving one.

Sonnet 5 just made the containment problem 10x worse by putting those capabilities on the default tier.

Audit every active Anthropic integration today.

Pin your model versions. Re-run your agent deployment readiness assessments. Update your AI model registry.

The EU AI Act conformity assessment obligations just changed for every workflow touching Sonnet 5.

Your vendor risk model just became your governance model.

SOURCE: https://www.anthropic.com/news/claude-sonnet-5
VERIFIED: Anthropic official blog (June 30, 2026), TechCrunch (June 30, 2026), AI Governance Institute (July 1, 2026)
SIGNAL: Enterprises just lost control of their AI deployment timeline. Every model provider can now push capability upgrades to default tiers without customer consent. This is the shadow AI problem on the vendor side.