Judge Says Pentagon Blacklist Looks Like “An Attempt to Cripple” Anthropic — Ruling Expected Within Days
After hearing oral arguments yesterday, U.S. District Judge Rita Lin told the government its Anthropic ban looks like punishment. “It looks like an attempt to cripple Anthropic,” she said from the bench, adding that the DOD’s actions appeared “troubling” because they weren’t narrowly tailored to the national security concerns cited — concerns that could have been addressed simply by ceasing to use Claude. Lin questioned whether the government violated the law when it designated Anthropic a supply-chain risk and when Defense Secretary Pete Hegseth announced that any company seeking Pentagon business must cut ties with Anthropic entirely.
The judge did not issue a ruling from the bench. She expects to file her written order on the preliminary injunction motion within days. If she grants it, the Pentagon designation would be suspended while the full lawsuit proceeds. Anthropic has told the court it faces billions in potential losses without interim relief. The government’s position — that refusing a contract clause is commercial conduct, not protected speech — appeared to get a skeptical reception.
Claude Code Auto Mode Ships — Safety Classifier Guards Every Action Before It Runs
Anthropic launched auto mode for Claude Code as a research preview. It’s a new permissions level that sits between the standard “ask for everything” mode and the risky “skip all prompts” flag. In auto mode, Claude makes permission decisions on your behalf — but before each tool call executes, a safety classifier reviews it for potentially destructive actions (mass file deletion, sensitive data exfiltration, malicious code), prompt injection attacks embedded in content Claude is reading, and anything else that looks unintended. Safe actions proceed automatically; flagged ones are blocked.
Auto mode is available now for Team plan subscribers using Claude Sonnet 4.6 or Opus 4.6, and is coming to Enterprise customers and API users in the next few days. Anthropic recommends running it in isolated environments during the preview period. The classifier may occasionally flag harmless actions or, in rare cases, allow risky ones when context is ambiguous — the company is iterating based on feedback.
Claude Code Channels: Text Your AI Coder from Telegram or Discord, Get Replies When It’s Done
Claude Code Channels, launched last week as a research preview, lets developers push messages from Telegram or Discord directly into a running Claude Code session on their local machine. Claude processes the request with full filesystem, MCP tool, and git access, then replies through the same messaging app when it finishes. No open ports, no tunneling — the integration runs as an MCP plugin locally. Anthropic built it on top of the same Model Context Protocol open standard it released in 2024, which is why adding new channel types (Slack, WhatsApp) is straightforward for the community.
VentureBeat described the release as Anthropic’s direct answer to OpenClaw, the open-source tool that had popularized asynchronous AI coding workflows via messaging apps. The difference: Claude Code Channels ships with Anthropic’s safety reputation and full Claude Code capability, not just a thin wrapper. Requires Claude Code v2.1.80+ and an active Team or Enterprise subscription.
SDK Tool Helpers in Beta, Search Content Blocks, and Free Code Execution with Web Search
The Python and TypeScript SDKs now include tool helpers in beta, simplifying tool creation and execution with type-safe input validation and an automated tool runner that handles multi-turn tool conversations. Alongside this, search result content blocks are in beta — tools can now return search results with source attribution, and Claude will automatically cite those sources in responses. This is a meaningful improvement for RAG applications that previously required custom citation pipelines.
On pricing: sandboxed code execution is now free when used alongside web search or web fetch, and both web search and web fetch now support dynamic filtering via code execution — results are filtered before hitting the context window, improving quality and reducing token costs. The Models API, Message Batches API, and Token Counting API are all now generally available (no beta header required). Console users with the Developer role can also now access the Cost page.
2x Usage Boost Ends March 27 — 2 Days Left to Maximize Off-Peak Limits
Anthropic’s March off-peak promotion expires at the end of March 27. Free, Pro, Max, and Team subscribers get double usage limits outside 8am–2pm ET on weekdays, and all day on weekends. The boost applies across Claude on web, desktop, mobile, Cowork, Claude Code, Claude for Excel, and Claude for PowerPoint. Enterprise customers are excluded. Capacity doesn’t roll over — if you have intensive workloads to run, this weekend is the window.
Anthropic IPO Watch: Wilson Sonsini Hired, $350B Round in Sight
Coverage of Anthropic’s IPO preparations continues to intensify. The company has retained Wilson Sonsini as its legal adviser and held preliminary discussions with investment banks. Current private-market valuation sits above $183 billion, with a new funding round reportedly targeting $350 billion. CEO Dario Amodei has projected annualized revenue will double to roughly $26 billion, with more than 300,000 business customers. No timeline for a public offering has been confirmed. The outcome of the Pentagon injunction — now expected within days — could materially influence IPO timing and structure.
Anthropic Is Building Two Different Kinds of Leverage Simultaneously
The Pentagon case and the product roadmap are telling the same story from different angles. In court, Judge Lin’s skepticism toward the government is the best possible signal Anthropic could have gotten short of a ruling. But the more interesting thing is that Anthropic seems to have decided it doesn’t need the Pentagon to win.
Think about what shipped in the last five days: computer use on macOS, auto mode for Claude Code, Channels for Telegram and Discord, and a string of API improvements. Each one extends Claude’s reach into a workflow where it previously couldn’t go. The Ramp data from Monday showed the commercial leverage that already exists — nearly 1 in 4 businesses, 70% of first-time buyer matchups won. Now Anthropic is adding autonomous desktop control, async coding from your phone, and safety guardrails that remove the friction from running long agentic tasks. The product surface is expanding fast enough that losing government contracts, while painful, starts to look survivable. That’s exactly the kind of leverage you want heading into both an IPO and a court ruling.