FTC Investigates OpenAI and Anthropic

Highlight

Claude Sonnet 5.5 becomes the default Sonnet model

Sonnet 5.5 is now the default Sonnet on the Anthropic API, with a 1M-token context window and pricing of $2/$10 per million tokens plus $0.20 per million for cache reads. If you run Sonnet for subagents or day-to-day coding, you get the larger context without a [1m] suffix, and /model will resolve the sonnet alias to it.

Worth Knowing

Claude Mods let plugins change core behavior

Mods are TypeScript functions that hook into Claude Code itself: you can rewrite prompts, block risky commands, draw custom panes, or replace built-in features, then ship the result as a plugin. A built-in example, You should know, runs a side agent that flags things you or Claude might miss, and you can turn it on with /plugin enable cc-plugin-you-should-know@builtin.

Sessions now start in auto mode by default

Interactive terminal and VS Code sessions start in auto mode when you have no permission mode configured, on every plan and provider, and claude -p and Python SDK sessions follow the same rule. Setting permissions.defaultMode or passing --permission-mode still overrides it.

1M context becomes the default on Bedrock and Vertex

Opus 4.7 and later plus Fable now default to a 1M-token context on Bedrock, Vertex, Foundry, the Claude apps gateway, and behind a custom ANTHROPIC_BASE_URL, with no [1m] suffix to pick. If your gateway stops at 200K, set CLAUDE_CODE_DISABLE_1M_CONTEXT=1 or run /autocompact 200k.

Up arrow restores a prompt you cleared with Ctrl+C

If you wipe a draft with Ctrl+C, pressing Up on the empty prompt brings it back, including pasted text and images. New keys also help elsewhere: Ctrl+F finds a session by name in the agents view, and Alt+↑/↓ jumps between groups.

VS Code adds bookmarks and run in background

You can save Claude's responses to a Bookmarks side panel, move a running command or subagent to the background from the chat, and see background shell and Monitor output on their cards in the agent map. Stop and Escape now end only the current turn, leaving background agents running until you stop them individually.

New desktop launch and plugin configuration commands

claude --desktop opens the Claude desktop app on your current directory, or on a session with --continue / --resume <id>, and claude plugin configure <plugin> shows a plugin's options or saves values from stdin with --values-stdin. claude project purge is now claude purge, with the old name still working.

Under the Hood

Permission and redaction gaps closed

A dangerous rm inside a bash -c or sh -c script no longer runs unprompted under bypassPermissions or a shell allow rule, Bash deny and ask rules now catch commands behind environment variable prefixes under sandbox auto-allow, and logs and transcripts stop leaking parts of URL passwords and keys with invisible characters.

Faster Bash calls and lighter startup

Bash commands lost a subshell that ran before every invocation, startup builds only the settings schema your config actually uses, and large MCP results now use less memory and produce smaller session files. A brief freeze partway through long non-fullscreen sessions is also gone.

From Anthropic

In the News