← All topics

Anthropic news and Claude timeline

189 company and product stories · 92 issues


Anthropic develops Claude models and applications. Claude supports conversation and documents, Claude Code handles programming tasks, and developers and businesses use the API to integrate models into their workflows. This page follows products, research, enterprise adoption and capital-market developments.

Editorial review · Coverage through

Recent coverage

Selected milestones

Dates refer to our coverage. Each milestone retains the reported stage at that time. Follow the news timeline for subsequent developments.

  1. Vulnerability scanning for open-source projects

    Anthropic launches OSS Scanner, using its strongest models to scan open-source projects for free and sending reports to maintainers without human review

    Read original coverage ↗
  2. Public IPO registration filing

    Anthropic files for IPO with $16.3 billion in first-half revenue and over $600 billion in compute commitments

    Read original coverage ↗
  3. Expanded scientific partnerships

    Anthropic expands AI for Science through 20 institutional partners

    Read original coverage ↗

Financing and transaction coverage

Dates are our publication dates. Announcements and talks are not closing; different reports may concern the same transaction.

  1. · Anthropic · Undisclosed · Compute financing (in talks) · In talks, not closed
  2. · Anthropic · Offering size not established · IPO preparation (reported) · Reported
  3. · Anthropic · Share count, price and proceeds not set · Public U.S. IPO registration filing · Filing or placement

Coverage & briefs

Filter complete history · 216 entries

Ars Technica AI says Claude Code settings files may expose secrets on GitHub

According to Ars Technica AI, Apiiro researchers found that Claude Code settings files can leak into public GitHub repositories and expose thousands of secrets. The exposed data reportedly includes database passwords, API tokens and root logins. The issue stems from how Claude Code stores user approval for automated shell commands in a local .claude/settings.local.json file. Anthropic says that file is usually added to a developer’s global .gitignore, but mishandling can still leave it public.

AI-generated

GitHub Copilot makes Anthropic’s Claude Fable 5.2 generally available

GitHub Copilot said on its blog that Claude Fable 5.2 is now generally available to Pro+, Max, Business and Enterprise users. The model is intended for complex, multistep coding across large codebases, and early testing found that it solved more tasks than Claude Fable 5.1 while using fewer steps and tool calls. It is available in several Copilot surfaces, including VS Code, Visual Studio, the Copilot CLI, GitHub.com, mobile, JetBrains IDEs, Xcode and Eclipse. Business and Enterprise admins must enable it manually, and data retention is required by default.

AI-generated

According to IT之家, Anthropic study says scientists use Claude most for analysis and writing

According to IT之家, Anthropic released a research report based on 207,000 research sessions from Aug. 2 to Sept. 30, 2026. The study found that scientists use Claude mainly for data analysis and writing, while data collection, research ideation, and hypothesis generation are less common. Users in computer science, math, and statistics were more active, and Claude Code was even more concentrated in quantitative research. Anthropic also surveyed 528 scientists, and the results broadly matched the session data.

AI-generated

According to the Financial Times, Anthropic takes over a $3.4 billion gas generator order

According to the Financial Times, Anthropic in late September spent $3.4 billion to take over a heavy gas-generator order from OnePWR Solutions in San Antonio, Texas, with total capacity of 1.7 GW. The report says it is the largest power-equipment purchase yet known for an AI company. As turbine and generator lead times stretch, Anthropic chose to bypass intermediaries and secure power capacity directly, with some units due soon and the rest over the next two years.

AI-generated

According to The Verge AI, METR says Anthropic’s AI is accelerating its own R&D

According to The Verge AI, AI safety research group METR found after an 11-week investigation that Anthropic’s models are improving roughly twice as fast as a year ago, amounting to about two years of historical progress in the last year. The report says Anthropic employees using its most capable internal model self-reported about a 12x productivity boost, and METR concluded most of the acceleration “is due to AI” rather than just more compute. METR also said Anthropic heavily redacted the public version of the report.

AI-generated

TechCrunch reports Anthropic adds a personal finance tool to Claude

According to TechCrunch, Anthropic announced a new Claude feature called Personal Finance that lets users connect bank and credit card accounts through Plaid and ask questions about balances, transactions, and spending. Anthropic says the access is read-only, so Claude cannot move, manage, or invest money, and chats in the finance hub are not used to train its models. The beta rollout begins this week for U.S. users on the Pro, Max, and Ultra plans.

AI-generated

Business Insider says Catherine Wu joins OpenAI after leaving Anthropic

According to Business Insider, Catherine Wu has joined OpenAI as product lead for its core products team after leaving Anthropic last week. She announced the move on X and said she has a lot of respect for OpenAI and its reach to more than 1 billion users. Wu previously led Claude Code, helping it grow from an internal prototype into a major commercial product. OpenAI confirmed the hire, and she will work with Nick Turley on enterprise products and Ashley Alexander on consumer products while reporting to Tibo Sottiaux.

AI-generated

Anthropic releases FlashProve to verify which model produced a response

Anthropic introduced FlashProve, a prototype that lets a provider prove a response came from the model it committed to using without disclosing its weights. In a test with a 4-bit version of Llama 4 Maverick, a single H100 generated proofs at around 110 tokens per second. The proof for a 1,024-token request was about 19 MB. Anthropic estimates that generating the proof took roughly 70 to 95 times the GPU time required to run the model itself with vLLM.

AI-generated
Mentions in other coverage · 147
SpaceX AI revenue reaches $7.56 billion, exceeding half of quarterly sales
AI revenue can come from models, software or rented compute; evaluating commercialization requires separating those streams from the capital needed to expand.
Musk pledges one SpaceX share per US child in 2027, with details pending
Equity-giving pledges should be judged by distribution rules, transparent execution and the rights recipients actually receive.
DeepSeek opens Search and Fetch APIs at RMB 5 per 1,000 successful requests, with existing tool calls billed from November
An agent's total cost increasingly depends on the tools it invokes and how often it invokes them, beyond the model's token prices.
Amazon is in talks to buy AI software company Decart for about $5.5 billion
The in-house chip race comes down to software: whoever makes models run smoothly on its chips can pull customers away from Nvidia.
North.cloud raises $60 million to help companies see where their cloud and AI money goes
The more AI companies use, the more they need to know where the money goes, and helping them save on AI has become an AI business itself.
DeepSeek merges four teams into a Product Department led by Harness chief Cui Tianyi
As gaps between models narrow, turning a model into a product people enjoy using is becoming the new dividing line between labs.
OpenAI reportedly trains staff in counterespionage, warning of honey traps, bribes and eavesdropping in public places
Once model companies' secrets are worth the effort of stealing, AI labs' security practices start to converge with those of defense contractors and intelligence agencies.
The X Subscription bundles Grok, Cursor and X benefits, with Grok Bot starting at $30 a month
AI subscriptions are spanning chat, execution and coding; shared allowances can strengthen loyalty, but savings depend on whose models a customer actually uses.
Jack Clark moves his full recursive self-improvement forecast to 2027, pointing to better research judgment
AI research automation is increasingly about designing useful experiments; benchmark gains and full recursive improvement remain distinct levels of evidence.
Amazon in talks to buy AI startup Decart for about $7 billion, which could give its in-house chips a software layer
Building a custom chip is only the first step; whether customers move depends on how painful migration is, and cloud providers are now paying heavily for the moving tools.
AI cloud provider Lambda is in talks for a pre-IPO round of up to $4 billion at a $14.5 billion pre-money valuation, after its backlog grew to $50 billion in three months
Compute-leasing companies are increasingly valued on the long-term contracts they hold, and with those contracts concentrated in a few model companies, their fates are tied together.
OpenAI tells investors annualized revenue is about $50 billion, below the $68 billion figure that had circulated, and Nvidia, Oracle and CoreWeave shares fall
The two leading model companies are hard to compare because nobody agrees whether sales made through cloud providers count as their own; prospectuses will force them onto the same arithmetic.
Model leaderboard Arena raises a $200 million Series B at a $3.1 billion valuation and releases an Alignment Index measuring whether models behave
An evaluation company more than tripling revenue in a year shows vendor-published scores are no longer enough to buy on, and refereeing is turning into a business of its own.
Broadcom seeks more than $50 billion in financing for OpenAI's custom chips, with Apollo and Blackstone among those in talks
From Nvidia to Broadcom, chip sellers increasingly double as their customers' financing arrangers, and it is getting hard to tell how much AI compute demand is being held up by suppliers' own credit.
Mistral releases its trillion-parameter Large 4, which third-party testing calls the strongest model from outside the US and China
Open models from Europe and America are both measuring themselves against Chinese ones, so on the open-weight track there is no longer any dispute about who leads.
SpaceX seeks $40 billion of debt, led by Apollo, to buy Nvidia chips
More and more of Nvidia's orders are being paid for with customers' debt, so the strength of chip demand is now tied to how much credit markets are willing to lend.
OpenAI is in talks for a new $30 billion round, with UAE funds discussing up to $10 billion in total
When a company can raise $30 billion at a stated price with no lead investor, what is scarce in capital markets is no longer money but assets large enough to hold it.
SpaceX stock rises 7.6% in a day to its highest since mid-June, returning Musk to trillionaire status
A rocket company's share price now needs its AI business to explain it, which shows markets are ready to reprice any asset that can be tied to AI demand.
Google is found to be developing "Superprojects" for Gemini, and a Nano Banana 2.1 label appears in Flow
Every assistant is competing for the "project" container, because whoever holds a user's files, email and calendar becomes the default place to work.
Musk says SpaceXAI will be renamed SpaceXSI, the unit's third name in nine months
An executive order cannot change what an industry calls itself, but it can make the companies that need the government most change their wording first.
AI agent security company Reco adds $55 million as customer AT&T becomes a strategic investor
Companies are deploying agents faster than they can count them, so the first thing security vendors sell is not protection but an inventory.
Nine-person Halluminate raises a $30 million Series A selling finance-task training environments to top labs
For models to learn the work of bankers and private-equity analysts, what is scarce is not compute but exam questions written by people who know the job, so the small firm selling the questions made money before the financial firms using the models.
MiniMax's M3.1-Flash-Preview is used to reproduce Opus 5.5's front-end showpieces, as flagship skills reach lightweight models faster
A flagship's signature skill showed up in a lightweight model within a week, so the window in which a leader can charge a premium for something only it can do is now measured in days.
SpaceXAI plans to bring about 420,000 Nvidia GPUs online in November and has discussed renting compute to Microsoft
Even the most aggressive builder of its own compute has found that selling compute makes the economics easier to justify than only using it.
Amazon discusses moving $8 billion of Nvidia chips into an SPV and leasing them back to expand its data centers
When the richest cloud giant starts moving chips off its balance sheet, the bottleneck in AI infrastructure has shifted from getting chips to making the financing math work.
Artificial Analysis: Gemini 4 Argon matches GPT-6 Astra on intelligence with a hallucination rate of just 15%
As frontier models' intelligence scores converge, knowing what they don't know is becoming the metric that separates them.
Google releases Gemini 4 Argon, which ties GPT-6 Astra in independent tests with under a third of its hallucination rate
When overall scores tie, saying fewer wrong things may win over high-paying enterprise customers more than getting one more question right.
OpenAI changes Pro plan usage math, and heavy developers say it halves their allowance
The hardest part of subscription AI isn't setting prices but changing them: once an allowance is given, taking it back burns the trust of your most important users.
Meta's Muse passes 3 million weekly prompting users, giving consumer agents a first stable base
Agents won't be won on download speed but on whether users think of them every day, a habit no company has built yet.
SpaceXAI plans a unified Grok and X subscription topping out at $100 a month
Bundling can quickly lift paid numbers, but people pay for social media and for AI for different reasons, and combining them won't necessarily make both more valuable.
// related topics — frequently covered alongside #Anthropic