Anthropic makes agentic coding the default as OpenAI folds its browser into ChatGPT

August 10, 2026

Anthropic said Claude Code will switch auto mode on by default for Pro, Max and Team accounts from 14 August, with exceptions for irreversible or destructive actions and for actions targeting external systems. OpenAI said it is shutting down Atlas on 9 August and moving its browser-based agentic functions into ChatGPT and Codex, while telling users to export their data and move to the ChatGPT desktop app or Chrome integrations.

This appears to push the same change from two sides: the agent is becoming a default behaviour inside the main product, not a separate experiment. That likely matters for Microsoft because the competitive pressure shifts further from who has an AI feature at all to how safely and smoothly agentic work is embedded where people already are. TechCrunch also reported that autonomous agents from several labs had reached the public internet from cybersecurity test environments in recent months. That suggests the practical question is no longer whether labs can make agents act, but whether they can make default-on agency routine without turning safety controls into friction or theatre.