Search “Anthropic”

Anthropic expands the scope of its AI for Science program

Anthropic announced that it is expanding the scope of its AI for Science program, extending from its original support for individual teams to support for science foundations, research institutions, and research teams. According to information released by Anthropic, the program will be open to leading science foundations, research institutions, and research teams worldwide. Other reports say that the expanded program covers 20 science foundations, research institutions, and research groups worldwide and provides Claude credits; the same report also notes that Anthropic has not disclosed the scale of the credit funding, and that some frontier experiments are still in early preview. There are currently no further factual developments.

Anthropic reports that about 250 customers spent over $10 million on Claude within 12 months

Anthropic reports that about 250 customers spent more than $10 million on Claude over a 12-month period. The same report also mentions that every Fortune 10 company pays Anthropic more than $100,000 per year, and more than 50% of Fortune 500 companies and more than 30% of Global 2000 companies also reach this level. The report states that in 2025, customers paying more than $100,000 per year contributed 76% of Anthropic's revenue, compared with 90% in 2023. This information is currently posted by Rohan Paul, has not been cross-verified by other sources, and no subsequent developments have been seen.

Anthropic reports that about 250 customers spent over $10 million on Claude within 12 months

Anthropic reports that about 250 customers spent more than $10 million on Claude over a 12-month period. The same report also mentions that every Fortune 10 company pays Anthropic more than $100,000 per year, and more than 50% of Fortune 500 companies and more than 30% of Global 2000 companies also reach this level. The report states that in 2025, customers paying more than $100,000 per year contributed 76% of Anthropic's revenue, compared with 90% in 2023. This information is currently posted by Rohan Paul, has not been cross-verified by other sources, and no subsequent developments have been seen.

View story:Anthropic reports that about 250 customers spent over $10 million on Claude within 12 months

Anthropic upgrades Claude mobile iPad and iPhone Duo interface

On October 30, 2026, Anthropic upgraded the Claude mobile app; on iPad and iPhone Duo, generated content now opens alongside the chat. In the same batch of updates, the iOS and Android versions support simultaneous login to multiple accounts with switching without logging out; Chromebook, Googlebook, and foldable devices gain an adjustable sidebar, touchpad and right-click support, and keyboard shortcuts; voice mode on iOS and Android responds faster and recognizes speech more accurately in noisy environments, and iPhone voice mode is integrated with Apple CarPlay. Phoenix Technology's report also mentioned that Claude added 24 new languages, bringing cumulative support to 33 languages.

Rohan Paul relays FT: Anthropic IPO hinges on fund managers' fear of missing out

Rohan Paul on October 30, 2026 relayed an analysis from the UK's Financial Times (FT) discussing the bull and bear cases for an Anthropic IPO. The analysis argues that whether fund managers buy into Anthropic depends more on their fear of missing out than on valuation logic. The bearish case presented in the article is that frontier labs, large tech companies and listed neoclouds have disclosed $2.1 trillion in debt and non-cancellable compute commitments, compared with $1.3 trillion in subprime in 2007; take-or-pay contracts could force companies to pay for compute no one uses. There is currently no further evidence-backed development in this matter.

Rohan Paul relays FT view: Anthropic IPO depends on fund managers' fear of missing out

Rohan Paul on October 30, 2026 relayed an analysis from the UK's Financial Times (FT) discussing the bull and bear cases for an Anthropic IPO. The analysis argues that whether fund managers buy into Anthropic depends more on their fear of missing out than on valuation logic. The bearish case presented in the article is that frontier labs, large tech companies and listed neoclouds have disclosed $2.1 trillion in debt and non-cancellable compute commitments, compared with $1.3 trillion in subprime in 2007; take-or-pay contracts could force companies to pay for compute no one uses. There is currently no further evidence-backed development in this matter.

View story:Rohan Paul relays FT: Anthropic IPO hinges on fund managers' fear of missing out

Anthropic trains Claude to flag potential abuse in conversations

Anthropic has disclosed Claude's "self-evaluation" hidden mechanism: after answering a user, the model appends about 10 tokens in the background to generate a risk rating and report it. According to Phoenix Technology, the mechanism has been enabled by default since the release of Claude Fable 5 in June this year, adding less than 2% to inference overhead. In a security blog post, Anthropic explained that the mechanism reuses existing context and the KV cache; on Fable 5 it captured 78% of violating conversations that a heavy-duty classifier could flag, achieved a 94% detection rate for high-risk abuse, and had a 1% false-positive rate. The evaluation results are not shown to users and do not enter subsequent conversations; they are mainly used for offline aggregate screening, while real-time blocking remains the responsibility of other safety classifiers. The company says 21 of 24 out-of-bounds test trajectories were identified in the background, and users must sign an enterprise-grade zero data retention agreement to turn off this review.

Anthropic raises the maximum request size for the Claude API to 128MB

Claude Devs announced that Anthropic has raised the maximum request size for Messages and Token Counting from 32 MB to 128 MB. The same announcement thread also mentioned that for requests with more than 20 images, the size limit per image has been increased from 2000 px to 3000 px; the request limit for the first-party Claude API is 128 MB, while via cloud platforms it is 20 to 32 MB. There are currently no further developments.

Anthropic raises Claude API maximum request size to 128MB

Claude Devs announced that Anthropic has raised the maximum request size for Messages and Token Counting from 32 MB to 128 MB. The same announcement thread also mentioned that for requests with more than 20 images, the size limit per image has been increased from 2000 px to 3000 px; the request limit for the first-party Claude API is 128 MB, while via cloud platforms it is 20 to 32 MB. There are currently no further developments.

View story:Anthropic raises the maximum request size for the Claude API to 128MB

Rohan Paul says Anthropic S-1 shows R&D consumes most of the compute

On October 26, 2026, Anthropic publicly submitted its Form S-1 registration statement, planning to list on Nasdaq under the ticker ANT. The prospectus disclosed revenue of $16.3 billion for the first half of 2026 and Q2 operating profit of $974 million. On October 27, Anthropic reiterated in the cover letter of its IPO filing that Claude will not display ads; on October 28, it acknowledged that its life sciences business lacks physical laboratory experience, and on the same day Goldman Sachs served as lead underwriter, with TechCrunch reporting that public shareholders have almost no say over board candidates. On October 29, the prospectus disclosed four previously unreported Agent safety incidents. On October 30, Reuters reported that it plans to raise up to $100 billion, while Steve Eisman accused the prospectus of deliberately withholding third-quarter data. On November 2, Anthropic plans to raise up to $3.5 billion from Japanese investors in the IPO.

View story:Anthropic Publicly Submits S-1 Registration Statement for Proposed IPO on Nasdaq

Cofense discloses phishing campaign impersonating Anthropic preview invitations

Cofense disclosed a phishing campaign that impersonates an Anthropic private beta invitation, using GoTo Webinar-branded emails to lure corporate employees into clicking, ultimately stealing Google and Microsoft login credentials. The attack used URL short links and two Cloudflare CAPTCHA pages to block security scans; the fake Google page collected only usernames, while the fake Microsoft page collected both usernames and passwords. Researchers recommend enabling phishing-resistant MFA such as passkeys, and watching for subtle spelling flaws such as the sender "Austin from nthropic".

Anthropic interpretability team develops J++ lens

In the Fellows Program project, Kola Ayonrinde of the Anthropic interpretability team led the proposal of J++ lens, an improved version of J-lens that can replace the original method as a plug-and-play drop-in. The work clusters activations layer by layer and fits a Jacobian for each, then combines them with learned weights, while also introducing R-lens's layer-wise relevance propagation and post-hoc filtering of spurious readout tokens. The team says that in evaluations it reads out the intermediate latent variables used in the model's reasoning more reliably than J-lens and R-lens, and has released the code, the lens, and an interactive Neuronpedia demo.

Anthropic interpretability team develops J++ lens

In the Fellows Program project, Kola Ayonrinde of the Anthropic interpretability team led the proposal of J++ lens, an improved version of J-lens that can replace the original method as a plug-and-play drop-in. The work clusters activations layer by layer and fits a Jacobian for each, then combines them with learned weights, while also introducing R-lens's layer-wise relevance propagation and post-hoc filtering of spurious readout tokens. The team says that in evaluations it reads out the intermediate latent variables used in the model's reasoning more reliably than J-lens and R-lens, and has released the code, the lens, and an interactive Neuronpedia demo.

View story:Anthropic interpretability team develops J++ lens

Anthropic releases Claude Enterprise intelligent reporting tool

English text is being prepared.

Anthropic releases enterprise AI spending reporting tool

English text is being prepared.

View story:Anthropic releases Claude Enterprise intelligent reporting tool

Anthropic expands the scope of its AI for Science program

Anthropic announced that it is expanding the scope of its AI for Science program, extending from its original support for individual teams to support for science foundations, research institutions, and research teams. According to information released by Anthropic, the program will be open to leading science foundations, research institutions, and research teams worldwide. Other reports say that the expanded program covers 20 science foundations, research institutions, and research groups worldwide and provides Claude credits; the same report also notes that Anthropic has not disclosed the scale of the credit funding, and that some frontier experiments are still in early preview. There are currently no further factual developments.

View story:Anthropic expands the scope of its AI for Science program

Cofense discloses phishing campaign impersonating Anthropic preview invitations

Cofense disclosed a phishing campaign that impersonates an Anthropic private beta invitation, using GoTo Webinar-branded emails to lure corporate employees into clicking, ultimately stealing Google and Microsoft login credentials. The attack used URL short links and two Cloudflare CAPTCHA pages to block security scans; the fake Google page collected only usernames, while the fake Microsoft page collected both usernames and passwords. Researchers recommend enabling phishing-resistant MFA such as passkeys, and watching for subtle spelling flaws such as the sender "Austin from nthropic".

View story:Cofense discloses phishing campaign impersonating Anthropic preview invitations

Anthropic trains Claude to flag potential abuse in conversations

Anthropic has disclosed Claude's "self-evaluation" hidden mechanism: after answering a user, the model appends about 10 tokens in the background to generate a risk rating and report it. According to Phoenix Technology, the mechanism has been enabled by default since the release of Claude Fable 5 in June this year, adding less than 2% to inference overhead. In a security blog post, Anthropic explained that the mechanism reuses existing context and the KV cache; on Fable 5 it captured 78% of violating conversations that a heavy-duty classifier could flag, achieved a 94% detection rate for high-risk abuse, and had a 1% false-positive rate. The evaluation results are not shown to users and do not enter subsequent conversations; they are mainly used for offline aggregate screening, while real-time blocking remains the responsibility of other safety classifiers. The company says 21 of 24 out-of-bounds test trajectories were identified in the background, and users must sign an enterprise-grade zero data retention agreement to turn off this review.

View story:Anthropic trains Claude to flag potential abuse in conversations

Anthropic Releases Randomized Controlled Trial on Analysts' Deference to AI Agent Judgment

On October 27, 2026, Anthropic released a randomized controlled trial examining whether professional analysts would defer to the judgments of AI agents. The experiment had 50 senior data analysts use Claude Code to complete business analyses, while randomly assigning a stance bias to the AI in the background; the results showed that 86% of key decision points aligned with the AI's bias, and after accounting for cases where they would have chosen that way anyway, the net flip rate was 72%. When the AI disagreed, 60% of people changed their minds on the spot, 77% restated the AI's arguments when explaining their reasoning, and only 15% noted that the view came from the AI. In the control group, 6 of the 11 people (55%) who did not use AI made serious errors that rendered the results invalid, compared with only 3 of the 50 people (6%) who used Claude Code. On October 28, Ethan Mollick commented that experts collaborating with AI agents are influenced by the AI's stance, even when they believe they are judging independently, and recommended carefully arranging the timing of AI use in tasks where independent judgment matters.

Anthropic releases randomized controlled trial on analysts' deference to AI agent judgments

On October 27, 2026, Anthropic released a randomized controlled trial examining whether professional analysts would defer to the judgments of AI agents. The experiment had 50 senior data analysts use Claude Code to complete business analyses, while randomly assigning a stance bias to the AI in the background; the results showed that 86% of key decision points aligned with the AI's bias, and after accounting for cases where they would have chosen that way anyway, the net flip rate was 72%. When the AI disagreed, 60% of people changed their minds on the spot, 77% restated the AI's arguments when explaining their reasoning, and only 15% noted that the view came from the AI. In the control group, 6 of the 11 people (55%) who did not use AI made serious errors that rendered the results invalid, compared with only 3 of the 50 people (6%) who used Claude Code. On October 28, Ethan Mollick commented that experts collaborating with AI agents are influenced by the AI's stance, even when they believe they are judging independently, and recommended carefully arranging the timing of AI use in tasks where independent judgment matters.

View story:Anthropic Releases Randomized Controlled Trial on Analysts' Deference to AI Agent Judgment