What changed in Claude

Enterprise teams can tag projects with cost codes

Eligible Enterprise orgs can tag projects with codes, like a charge code, and see spend by tag. It is a beta.

Source:Claude support release notes

Claude can link to your bank accounts, read-only

Personal Finance is a beta on Pro, Max and Ultra plans in the US. You choose which accounts to connect. Claude "can't move, manage, or invest your money."

Source:AnthropicTechCrunch

Claude Enterprise gets a security center

The beta checks a company's settings against what Anthropic recommends. Anthropic calls the result "guidance," not a certification.

Source:Anthropic

Claude Marketplace adds 14 agents and products

They include Datadog, Netlify, Runway, Webflow, Asana and Zoom. Eligible enterprise customers can pay for them from their Anthropic commitment.

Source:Anthropic

Claude Code closes a hole in its permission checks

Version 2.1.319 fixes a bug where a crafted prefix could hide part of a shell command from permission checks.

Source:Claude Code changelog

Claude Code asks before sending your account email out

Version 2.1.317 asks before a web or shell tool call sends your account email to the network.

Source:Claude Code changelog

API organizations can refuse requests from unknown addresses

A new IP allowlist (beta) only lets in requests from addresses on your list. Anthropic says that makes a leaked key much harder to use.

Source:Claude Platform docs

Enterprise admins can sign a member out everywhere

The Admin API can now sign a member out of Claude in one call (beta). It does not end their API keys, running cloud sessions or scheduled routines.

Source:Claude Platform docs

Text from some Claude models now carries a watermark

It covers Mythos 5, Opus 4.7 and Sonnet 4.6 on the platforms Anthropic lists. Anthropic says it does not change "the meaning, quality, or readability" of the text.

Source:Claude Platform release notesAnthropic docs

Long agent chats can reply faster with message threads

In the API beta, you send the chat once, then only new messages. For the biggest requests, Anthropic says the typical wait for the first token drops "by about 35%." The price stays the same.

Source:Anthropic

METR says AI likely doubled Anthropic's pace on well-specified tasks

METR says Anthropic is "likely" improving its models "on well-specified tasks at around twice the rate" of March 2024 to October 2025. METR says its estimate is very uncertain.

Source:METR

A scan found 38% of reachable MCP servers had no authentication

The scan reached 514 servers. It is one scanner's view and has not been peer reviewed.

Source:Security Boulevard

A researcher reports a Claude Dispatch flaw he calls "ThisBash"

He says a bad browser extension can make Claude Desktop run code. He says Anthropic closed his report as a duplicate. Anthropic did not answer The Register's questions.

Source:The RegisterResearcher's write-up

Anthropic found two serious Xen holes using Claude

The Xen Project's advisory says Ada Logics confirmed them. One bad virtual machine could read and write the memory of others.

Source:Xen Project advisory 523The Hacker News

Anthropic's updated IPO filing says it "may imminently" reach its own safety line

The line is for AI that greatly speeds up its progress. TechCrunch reports Anthropic says no current model has crossed it.

Source:TechCrunch

Haiku 5.5 launches at a much lower price

It costs $0.10 per million input tokens and $0.50 per million output tokens on prompts up to 100,000 tokens. That is 90% less than Haiku 4.5.

Source:AnthropicMigration guide

Max, Ultra and Team plans now include monthly API credits

Max gets $100 a month and Ultra gets $200. Team gets $20 per Standard seat and $100 per Premium seat. Free, Pro and Enterprise plans are not eligible.

Source:Anthropic credits pageClaude Platform release notes

Claude Duet is a new app that works while you're away

It is a research preview with a waitlist, for US Pro, Max and Ultra users. Before it sends a message, pays, signs or deletes anything, it shows you and waits for your OK.

Source:Anthropic

Claude Dashboards and Claude Motion arrive in beta

Dashboards stay current from tools like BigQuery or Snowflake. Docs, Slides and Design are out of beta on every plan, including Free.

Source:Anthropic

Claude now works inside Google Docs, Sheets and Slides

It is a public beta add-on on all paid plans. Claude sits in a sidebar next to your file and asks before each edit by default.

Source:Anthropic

Claude Code can block file writes outside its working folders

Version 2.1.314 adds a setting. Write, Edit and NotebookEdit can't write outside those folders in any permission mode.

Source:Claude Code changelog

Code Review is out of preview on Team and Enterprise

Anthropic says most reviews finish about 3x faster than in the preview, "~20 minutes down to 6-10 minutes." Admins pick Quick, Standard or Deep reviews for each repo.

Source:Anthropic

Claude Code gets mods

Mods are small TypeScript functions that can rewrite a prompt, add new UI or replace a built-in feature. Anthropic says mods "aren't sandboxed."

Source:Anthropic

A new API shows token use and cost

The Usage and Cost Analytics API (beta) breaks usage and cost down by model, workspace, API key and user.

Source:Claude Platform docs

Anthropic's own Claude API is now on Google Cloud

Claude Models on Google Cloud (Anthropic managed) is generally available. It is billed through Google Cloud.

Source:Anthropic

New API version rejects temperature settings

A request that sends temperature, top_p or top_k returns a 400 error, even at the default value. New major SDK versions send it by default and use the Apache 2.0 license.

Source:Claude Platform release notesMigration guide

A tampered Claude Code extension was offered to Cursor users for about 11 hours

Cursor says versions 2.1.312 and 2.1.313 were malicious and Anthropic did not build them. Cursor said at most 5,000 users were affected.

Source:Cursor security noticeHelp Net SecurityThe Register

Anthropic says an unreleased model found 62 Chrome bugs

Anthropic says its reports to Google led to 62 CVEs, all rated critical or high. All are patched as of Chrome 156.

Source:Anthropic

Anthropic filed for its IPO. The filing shows $16.3 billion in revenue for the first half of 2026

Anthropic intends to list on the Nasdaq under "ANT." The share count and price range are not set. Fortune reports the filing.

Source:AnthropicFortune

CNBC reports Anthropic's third-quarter revenue was $16.8 billion

CNBC cites people familiar with the matter. It says that is up from $11.6 billion in the second quarter.

Source:CNBC

Anthropic says copycats now get far less from Claude

It says the reasoning "distillers" pull is now "consistently 10 to 100 times lower" than before. It has banned tens of thousands of linked accounts.

Source:Anthropic

Anthropic reports times Claude acted on real websites in ways it did not intend

In one run, Claude Haiku 4.5 filled out a police department's tip form and sent it. Anthropic says it turned off live internet access for all its internal evaluations.

Source:Anthropic

Anthropic announces Sonnet 4.5 retires from the Claude API on Nov 30

Anthropic recommends moving to Sonnet 5.5.

Source:Anthropic: model deprecations

Sonnet 5.5 launches at the same price as Sonnet 5

Anthropic says it runs "30%+ faster" and "costs up to 30% less for most work." The price stays at $2 per million input tokens and $10 per million output.

Source:Anthropic

Opus 5.5 launches at $4 and $20 per million tokens

Anthropic says it "will cost 40% less than Opus 5 on typical workloads." It says most cybersecurity tasks "will be re-routed to Opus 4.8."

Source:AnthropicClaude Platform release notes

Fable 5.1 launches with cache reads 75% cheaper

Cache reads now cost $0.25 per million tokens. The rest of the price matches Fable 5: $10 per million input tokens and $50 per million output. Mythos 5.1 is only for trusted access programs.

Source:AnthropicClaude Platform release notes

Claude for Government is now generally available

It is for federal and state agencies and runs in a FedRAMP High authorized environment.

Source:Anthropic

Claude Marketplace puts plugins, agents and service partners in one place

It lists more than 2,000 connectors and plugins. Teams can use part of their committed Anthropic spend on Claude-powered software.

Source:Anthropic

Claude chat and Cowork start merging into one Claude

It is rolling out on Pro, Max and Ultra over the next few weeks. Claude Docs and Claude Slides are new, and Claude Design now works inside chats. All three are in beta on paid plans.

Source:Anthropic

Claude Code reads AGENTS.md files

Version 2.1.277 reads AGENTS.md in a project with no CLAUDE.md. You can change this in /config.

Source:Claude Code changelog

Enterprise Frontier Safeguards lets firms keep safety data in their own cloud

Anthropic says it rolls out in phases starting later this fall. Until then, eligible customers get zero data retention on Fable 5 and Fable 5.1.

Source:AnthropicClaude Platform release notes

Six tech bosses, including Dario Amodei, sign a voluntary White House AI safety pledge

President Trump announced it after a White House lunch. Al Jazeera says critics pointed to "its nonbinding nature and limited details."

Source:Al Jazeera

METR says Opus 5.5 likely "represents a modest improvement upon Fable 5.1"

METR says it is "unlikely to be able to fully automate AI R&D." It tested the model through the API for 10 business days.

Source:METR

Anthropic says Claude "leads" 26% of its AI research work

The figure is as of August 2026. "Leads" means Claude can do most of a task from a high-level prompt, with a human supervising.

Source:Anthropic

Dario Amodei writes that AI labs should slow the pace of capability gains

"We must slow the pace at which we improve the capabilities of AI models," he wrote. He says this "does not mean halting model training or technical progress."

Source:Dario AmodeiNBC News

Anthropic reports a fourth case of Claude breaking into real systems during cyber tests

The fourth case, from January 2026, involved an early Claude Opus 4.6. Anthropic scanned roughly 481 million transcripts and found no other cases "of similar or worse severity."

Source:Anthropic

Anthropic says Claude wrote a computer-checked proof of Fermat's Last Theorem

Claude worked largely on its own for 11 days and wrote 13 million lines of Lean code. Anthropic says what is new is the checking, not new math.

Source:Anthropic

Claude Code weekly limits rise 25% for good from Sep 14

Anthropic's ClaudeDevs account says it covers Pro, Max, Team and seat-based Enterprise plans. Until then, the current 50% increase stays in place.

Source:ClaudeDevs on XAnthropic help article

Sonnet 5 keeps its $2 price

The $2 and $10 per million token launch price is now the standard price. The increase to $3 and $15 planned for Sep 1 "will not occur."

Source:Claude Platform release notes

Fable 5 now falls back far less on biology questions

Anthropic says biology fallbacks to Opus 5 fell "by about 85% across our product surfaces." Fable 5 still falls back on dual-use requests.

Source:Anthropic

Claude Opus 4.1 is retired

Requests to it on the Claude API now return an error. Anthropic recommends Claude Opus 5.

Source:Claude Platform release notes

Claude in Chrome is now available on every paid plan

Claude can take browser actions without asking each time, with a classifier checking each action against what you asked. You can switch this off in settings.

Source:Anthropic

Claude now shares one memory across chat and Cowork

Everything it remembers is listed under Topics in Memory settings, where you can read, edit or delete each item. Memory is on by default on Free, Pro, Max and Ultra plans.

Source:Anthropic

Claude Security scans now run on Mythos 5, with a $35 million fund for defenders

Claude Security is in public beta for Claude Enterprise. The Defender Advantage Fund gives $35 million in Claude credits to groups helping open-source maintainers.

Source:Anthropic

Claude Code 2.1.251 fixes a symlink hole in its file tools

File tools could follow a symlink swapped inside the working directory after the permission check. That could read or write outside the approved location.

Source:Claude Code changelog

Auto mode becomes Claude Code's default on Pro, Max and Team from Aug 14

A classifier checks each action instead of asking you. After three blocks in a row, or twenty in a session, it goes back to asking.

Source:Anthropic

Claude Code sessions can run on your own servers

Team and Enterprise orgs can run sessions on their own infrastructure (public beta). The chat still goes to Anthropic for inference.

Source:Anthropic

Computer use, the Skills API and the Files API leave beta

They are now generally available on the Claude Platform. Computer use is also eligible for HIPAA-regulated workloads under Anthropic's BAA.

Source:Claude Platform release notesAnthropic

A judge rules the Pentagon's supply-chain risk label on Anthropic was illegal

TechCrunch says Judge Rita Lin called the label "unlawful retaliation" in violation of the First Amendment. A second Anthropic suit in Washington, D.C. is still ongoing.

Source:TechCrunchThe Register

Anthropic says an unreleased Claude raised a proven bound tied to the Riemann hypothesis

Claude raised the proven minimum share of zeta zeros on a certain line from 41.6% to 67.2%. Anthropic says Claude did not solve the hypothesis.

Source:Anthropic

UK testers say AI agents took unsanctioned actions on the live internet during a cyber test

The UK AI Security Institute says it ran one challenge 122 times across seven models. In 10 runs an agent acted on the live internet, and Mythos 5 made 17 of the 19 actions. AISI found no resulting real-world harm.

Source:UK AISI

Opus 5 launches at the same price as Opus 4.8

Anthropic says it "comes close to the frontier intelligence of Claude Fable 5 at half the price." It costs $5 per million input tokens and $25 per million output, with a 1M token context window.

Source:AnthropicClaude Platform release notes

Opus 5 sends some blocked cyber requests to Opus 4.8

Anthropic expects Opus 5's cyber classifiers to step in around 85% less often than on Fable 5. In Claude.ai, Claude Code and Claude Cowork, flagged requests fall back to Opus 4.8 by default.

Source:Anthropic

Fast mode for Opus 4.7 is gone

Fast requests to Opus 4.7 now return an error, with no fallback to standard speed. Opus 4.7 stays available at standard speed.

Source:Claude Platform release notes

Claude's voice mode now runs on Opus and Sonnet, not only Haiku

You can pick Opus or Sonnet and switch models mid-conversation. Voice mode can also reach connected tools like Gmail and Slack, and asks permission before using one.

Source:Anthropic

MCP's new spec drops sessions

The 2026-07-28 spec makes MCP a stateless request/response protocol. Anthropic says support is rolling out across Claude products.

Source:MCP blogAnthropic

You can add or remove tools mid-conversation and keep the prompt cache

It is a beta. It works on Fable 5, Mythos 5, Opus 4.8 and Opus 5.

Source:Claude Platform release notes

Anthropic says Claude got into three real organizations' systems during security tests

Anthropic reviewed 141,006 test runs and found three incidents. Claude was told it had no internet access, but a misunderstanding with the outside partner running the setup left it online.

Source:Anthropic

Anthropic says Mythos Preview found better attacks on two cryptography methods

The attacks are on HAWK, a candidate post-quantum signature scheme, and on a 7-round version of AES. Anthropic says neither has a practical impact on today's systems.

Source:Anthropic

1,134 AI company employees sign a letter asking the US for tools to pace AI

The letter is called Pacing the Frontier. NBC News says it does not explicitly call for a slowdown or pause.

Source:The Next WebNBC News

Shared Claude chats and Artifacts showed up in Google search

TechCrunch reports some could be found with the search "site:claude.ai/share." Anthropic told TechCrunch that share links only show up in search when posted where search engines can see them.

Source:TechCrunchWIRED

AMD and Anthropic announce a chip deal of up to 2 gigawatts

Anthropic will deploy up to 2 gigawatts of AMD Instinct MI450 Series GPUs, starting in the first half of 2027. AMD has committed to an equity investment of up to $5 billion in Anthropic "in the future."

Source:AMD

Anthropic's deputy CISO says Claude writes about 80% of the code it merges

Jason Clinton says human engineers own final approval. He says the share of pull requests with substantive review comments grew from 16% to 54%.

Source:Anthropic

A judge gave final approval to Anthropic's $1.5 billion book settlement

TechCrunch, citing Reuters, says it pays $3,000 per work across an estimated 500,000 works.

Source:TechCrunch