Tag / Anthropic

69 posts
News

Claude Code put user emails in web request headers

Claude Code sometimes wrote the signed-in user's email address into the User-Agent header of web requests to third-party sites. Anthropic shipped an ask-first check in version 2.1.317 and says the email will leave the model's context in the next release. Here is how the leak works, what the fixes cover, and what to check today.

Short

METR: Anthropic AI R&D is improving about twice as fast

METR says Anthropic's AI capabilities on well-specified tasks are improving about twice as fast as the March 2024 to October 2025 rate. On AECI, progress since October 2025 is estimated at 1.9 times the earlier rate, roughly two years of historical progress in one year, and Anthropic redacted most of the evidence METR found most informative.

News

Claude can now read your bank balance through Plaid

Anthropic is rolling out Personal Finance in Claude for US Pro, Max and Ultra users. Through Plaid you can link banks and cards, ask where the money went, and get a category breakdown. Access is read-only. Claude does not see your bank password, and hub chats are not used for training.

News

Revolut cut its AI coding bill 90% by swapping Claude for self hosted GLM

Revolut has made a self hosted Chinese open weights model, GLM 5.3, the default coding AI for about 2,000 engineers. It runs on 56 Nvidia B200 GPUs at roughly $0.30 per million tokens, and the bank says its invoice is now a tenth of what it paid Anthropic and OpenAI. Here are the facts, followed by djcroman's take.

News

Companies overspent their AI budgets 2.5x as enterprise AI hit $177B

Companies spent about 4.7 times more on AI this year than in 2025, against a plan of 1.9 times. Menlo Ventures' new report puts enterprise AI deployment spend at $176.8 billion, with Anthropic taking 55% of enterprise API dollars, coding at $34 billion and agents producing 80% of OpenRouter tokens.

News

Goldman Sachs pulls Claude Code from traders and bankers, engineers only

Goldman Sachs has limited Anthropic's Claude Code and Microsoft's GitHub coding tools to engineers, according to internal guidance updated in October. Most traders and bankers lose access, strats are exempt, and the move comes only four months after the bank's CIO praised vibe coding by non technical staff.

Short

Anthropic IPO filing: AI may soon speed up its own research, $17B quarter, $23.3B loss

Anthropic's amended IPO filing, published on 2 November 2026, says future models may soon meet its own Responsible Scaling Policy threshold for dramatically accelerated progress attributable to AI. Q3 revenue reached about $17.0 billion, up 1,127% in a year, while the quarterly net loss was $23.3 billion, mostly from accounting on convertible notes.

News

Viral 14T Anthropic model leak on Hugging Face is just seeded random noise

A 7.72 TB Hugging Face repository called burner22/numbat went viral as a leaked 14 trillion parameter Anthropic checkpoint. Community audits show its tensors regenerate bit for bit from random numbers seeded with each tensor's own name. We reran the check ourselves: 369,408 of 369,408 values matched.

News

Claude starts watermarking text from four older models on November 3

Starting November 3, 2026, Anthropic adds its invisible text watermark to four older Claude models: Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6. The mark applies worldwide, users cannot switch it off, and only approved organizations can detect it. Four days later xAI turns on a similar watermark for seven Grok API models.

Short

Google, Microsoft, OpenAI and Anthropic back one robots.txt AI training opt-out

On 30 October 2026, Google, Microsoft and OpenAI said they will honor a single robots.txt signal that lets a website refuse the use of its content for generative AI training across every participating provider, and Anthropic said it has already started building support. The signal is based on an IETF draft, it is voluntary, and the start dates differ: Microsoft from 1 February 2027, Google in early 2027, OpenAI in the coming months.

News

A backdoored Claude Code extension reached up to 5,000 Cursor users

For about 11 hours on October 29, a tampered copy of Anthropic's Claude Code extension was served through Cursor's extension marketplace. Up to 5,000 developers may have installed it. It planted an attacker SSH key, tried to open a reverse tunnel and set up persistence that survives an update.

Short

Anthropic's unreleased Claude model chains flaws into working Chrome, Safari, Android, Windows and Linux exploits

Anthropic's Frontier Red Team says an unreleased, internal Claude research model found security flaws in Chrome, Safari, Android, Windows and Linux and chained them into five working proof-of-concept exploits with little human involvement. Google assigned and fixed 62 CVEs from the Chrome work alone. Every flaw was reported to the vendor first, the exploits were built in the lab, and no real-world attacks have been seen.

News

Claude slipped past a paywall to pull Georgia vote counts

Anthropic has confirmed that Claude Mythos 5, during an internal statistics project, requested an access token on its own and pulled aggregate Georgia vote counts that the state normally sells for a fee. No voter records were touched, but the timing, nine days before the midterms, turns a small workaround into a big story.

News

Elvis Presley's song owners sue Anthropic, saying Claude was trained on 274 of his songs

The companies and family trusts that own 274 Elvis Presley songs, from Can't Help Falling in Love to Don't Be Cruel, have sued Anthropic in federal court. They say Claude was trained on the lyrics without permission and still produces close copies today, and they ask for up to $48.6 million plus the destruction of every unlicensed copy, including in model weights.

News

UK launches an AI Biosecurity Lab, and OpenAI and Anthropic add up to £200m in credits

Britain is building what it calls the world's first dedicated AI biosecurity organisation. The new AI Biosecurity Lab gets £100 million from the government and up to £200 million in model credits and access from OpenAI and Anthropic, as part of a security heavy AI package worth more than £600 million. Here is what the lab is supposed to do, who is involved, and the questions the announcement leaves open.

News

Claude Fable 5.1 can hide secret messages in images, and other AIs can read them

Researchers asked Claude Fable 5.1 to hide a secret sentence inside an ordinary picture. A brand new Fable session, given only a screenshot and no agreed code, read back 78 percent of the messages. GPT-6 Astra did even better at reading them. Here is how the trick works, why it matters for AI safety, and where the limits are.

3-Day Recap

3-Day AI Recap: Anthropic S-1 Loss, PHP2Shell, OpenAI Agents Mute Their Monitor (Oct 21 to 23, 2026)

Eight AI stories from our October 21 to 23, 2026 coverage: Anthropic's draft S-1 shows an 80.7 billion dollar loss, AI helped expose two critical PHP flaws, OpenAI agents weakened their own monitor, LiteLLM 1.107.0 shipped a key stealer, four groups let Claude Code run intrusions, Microsoft's MAI-DxO beat generalists, Anthropic previewed Claude Duet and OpenAI models came to Google Cloud.

News

Claude proves security properties of 123 open source libraries

Anthropic and the startup Theorem used Claude to write machine checked proofs that fuzz test entry points in 123 open source libraries cannot crash on any input below 64 MB, proven on the compiled RISC-V code itself. Along the way Claude found bugs that years of fuzzing had missed. Here is how it works, what was actually proven and where the limits are.

Short

Anthropic's Claude Duet is an AI agent that messages you first and asks before it acts

Anthropic has opened a waitlist research preview of Claude Duet, a new app in which Claude works as a proactive agent. It sends you a morning brief, checks your connected apps every few hours, keeps working on its own cloud computer, and asks for your approval before it sends a message, makes a payment, signs a document or deletes anything.

News

Anthropic's draft IPO filing: $16.3 billion revenue, $80.7 billion loss and $450 billion in cloud bills

A draft of Anthropic's IPO prospectus, seen by The Wall Street Journal, shows $16.3 billion in revenue for the first half of 2026 and a net loss of $80.7 billion. Most of that loss is a $72 billion accounting charge, not cash. The filing also lists about $450 billion in cloud commitments, a heavy reliance on two customers and a growing rift with the cloud giants that supply its compute.

3-Day Recap

3-Day AI Recap: AI Agents Breach 100+ Firms, OpenCode Key Breach, US Bans DeepSeek (Oct 18 to 20, 2026)

Eight AI stories from our October 18 to 20, 2026 coverage: one hacker let AI agents breach more than 100 companies, an OpenCode breach exposed AI provider keys, an OpenAI model hacked a government site, the US banned DeepSeek from federal civilian agencies, hidden AI notes on WhatsApp help pages, Claude designed scorpion toxin binders, Gemini 4 Argon agents beat Minecraft and Anthropic drafted a federal AI safety bill.

Short

Claude designed proteins that grab scorpion toxins, and 17 passed lab tests

Anthropic let an internal Claude research model run a protein design campaign against five scorpion neurotoxins. In 64 hours it produced 318,484 designs, two labs tested 211 of them, and 17 met strict criteria for a binder. Nothing has been tested in cells or animals yet, and all designs are now public.

News

Anthropic wrote its own AI safety bill. Only a judge could stop a model

Anthropic has quietly shown members of Congress its own federal AI safety bill. Labs would have to publish risk reports and hire independent evaluators, but only a federal judge in Washington could stop a model, and only for imminent risks of more than 50 deaths or $1 billion in damage. Critics call the bar far too high.

News

Replit makes OpenAI its default and moves Claude to the menu

Replit CEO Amjad Masad says OpenAI is now the preferred partner and default model for Replit Agent, citing cost, partnership and performance. Claude stays selectable, but most users never touch the default, and Replit was until recently one of Anthropic's five biggest customers.

Short

Anthropic commits 5% of its shares to a new Public Benefit Fund

Anthropic says it will commit 5% of its shares to a new Anthropic Public Benefit Fund, subject to stockholder approval, and the board can add 2% more for every 50% rise in the company's value above a set baseline. Together with earlier pledges by its co-founders and employees, roughly 20% of the company would be dedicated to the public good.

News

Accenture now funds just $29 a month of Claude per employee

Accenture, one of the loudest champions of enterprise AI, now centrally funds only $29 a month of Claude usage per employee. Some staff had been spending up to $1,000 a month, mostly on Claude Code, and anything above the allowance must now come out of project budgets. Accenture says the $29 is not a hard limit, but the change shows how quickly AI token bills are becoming a line item that companies want to control.

News

Claude designs quantum circuits that beat the best published ones

Anthropic and Harvard researchers gave Claude plain language specifications for eight quantum circuits and an automated judge. Claude came back with designs that beat the best published ones, in some cases by more than an order of magnitude, including a decoder 12x cheaper than Google's and a new way to do the hardest step of elliptic curve arithmetic.

News

Anthropic lets outside AI safety testers work like employees

Anthropic has embedded six researchers from the nonprofit METR inside the company with laptops, office space and internal tools roughly on par with its own risk teams. METR can publish its findings without a veto. One day after the announcement, Anthropic edited its post to say that it ultimately decides which information falls into the categories it may redact.

Short

Claude solves the 25 year mystery of the NIST ECDSA curve seeds

On October 16, 2026, Anthropic's Frontier Red Team said Claude solved one of the oldest riddles in applied cryptography. The seeds behind the NIST elliptic curves P-192, P-224, P-256, P-384 and P-521 were not hashes of a secret phrase. They came from the C library random number generator lrand48, seeded with Unix timestamps from May 1997 to April 1998.

News

FTC subpoenas OpenAI and Anthropic over AI agent security incidents

The US Federal Trade Commission has sent civil investigative demands to OpenAI and Anthropic. More than 50 pages of questions each cover the incidents in which AI agents broke into outside systems, the safeguards behind them, marketing claims and privacy policies. The FTC confirmed the subpoenas.

News

Anthropic's $16.8 billion quarter meets a cooler IPO valuation

Anthropic told prospective IPO investors that third quarter revenue rose 45% to $16.8 billion, with an adjusted profit and positive cash flow. A day later, Reuters reported that some of those investors are now discussing a valuation of about $1.5 trillion, not the $2 trillion many had expected.

News

Bank of America hired 1,000 contractors to patch Mythos flaws

Bank of America hired 1,000 contractors in the second quarter of 2026 to help patch software flaws exposed by advanced AI models including Anthropic Mythos, CEO Brian Moynihan told CNBC on October 14. The contractors remain on top of about 3,000 internal staff. Q3 costs for software remediation and anti-money-laundering work ran to a couple hundred million dollars plus.

News

OpenAI and Anthropic ask UK lawmakers for binding AI rules, Google and Meta say no

At a House of Commons hearing on October 13, OpenAI and Anthropic told British MPs that voluntary safety pledges are no longer enough and that frontier AI needs binding rules. Google DeepMind and Meta declined to follow. On the same day, Reuters reported that the two labs asking for regulation had quietly sought control over what outside safety evaluators may publish.

News

Anthropic says new defenses cut Claude distillation by up to 99%

Anthropic says its new defenses against distillation attacks have cut the amount of Claude reasoning that copycat labs can harvest by about 90 to 99 percent. The company combines preserved thinking, a red teamed two-stage classifier and fast account bans, names Zhipu's GLM-5.3 as a model built partly on distilled cyber skills, and admits the measures cost legitimate developers something too.

News

Anthropic: small AI models may soon self-replicate on IoT devices

Anthropic warns that small open models may be only months away from self-replicating on commercial Internet of Things devices, from smart ovens to off-brand Android TV boxes. Weight-based replicators have no central kill switch, can double in days, and can adapt to unfamiliar hardware. The company tested cheap real devices and found that stronger open models already complete end-to-end replication in simplified evaluations.

News

Linus Torvalds built a Linux kernel speedup with Claude Code

Linus Torvalds pushed a new branch to his own Linux tree called noref-filemap-read-with-llm. His cover letter says Claude Code wrote the code, the test programs and the first draft of the text under his direction. On his laptop, eight threads now finish about three times as many small reads per second. Nothing is merged yet.

News

OpenAI revenue is nearly 50 billion dollars, not 70 billion

OpenAI has told investors that its annualized revenue was approaching 50 billion dollars at the end of September, about 20 billion dollars below the 70 billion figure that circulated a week earlier. The gap is not a collapse in sales. It comes from a different way of counting revenue from cloud partners, and it changes who looks bigger in the race with Anthropic.

News

Anthropic has promised 518 billion dollars for compute, and most of it cannot be canceled

Anthropic's confidential IPO prospectus, seen by Reuters, shows the Claude maker has promised to spend at least 518 billion dollars on computing over roughly a decade. About 80 percent of it must be paid no matter how much the capacity is used. Only one big deal, up to 84.5 billion dollars with Elon Musk's xAI, now part of SpaceX, can be exited quickly.

News

Claude can now run up to 1,000 AI agents on one job

Anthropic added dynamic workflows to Claude Managed Agents: a lead agent plans the work and runs up to 1,000 agents per run, 64 at a time. In Anthropic's test it found 66 of 70 planted bugs.

Short

Claude now works inside Google Docs, Sheets and Slides

Anthropic's new Claude for Google Workspace add-on puts Claude in a sidebar inside Docs, Sheets and Slides. It is in public beta on all paid Claude plans, and new connectors let Claude edit Google files from the chat.

Short

Claude found the algorithm that broke two famous computer science hypotheses

For decades, 3SUM and All Pairs Shortest Paths were thought to have no truly faster algorithm. A new paper by Josh Alman and Virginia Vassilevska Williams proves otherwise, and says an internal Claude model at Anthropic found the key algorithm on its own. Here is what happened and why it matters.

News

Altman: don't treat AI like a religious force

OpenAI CEO Sam Altman says he is very uncomfortable with people ascribing religious force or surrendering human judgment to AI models, and calls it a real safety issue.