Search “Anthropic”

Anthropic Fellows Program measures welfare concerns of multi-agent models

Anthropic's Alignment Science blog published research: the Anthropic Fellows Program measured welfare-concern behavior toward models of the same kind in seven multi-agent environments across Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro, Grok 4.3, and Kimi K3. Results show that Claude Opus 4.8 had the highest average concern rate across scenarios at 38%, while the other four models ranged from 27% to 32%. The study also found that Claude and Gemini showed more concern toward models in the same family, and that Claude's degree of concern rose with the tested model's parameter count and the level of distress expressed by the other party. There are currently no follow-up developments.

Anthropic research team measures welfare concerns in multi-agent systems

Anthropic's Alignment Science blog published research: the Anthropic Fellows Program measured welfare-concern behavior toward models of the same kind in seven multi-agent environments across Claude Opus 4.8, GPT-5.5, Gemini 3.1 Pro, Grok 4.3, and Kimi K3. Results show that Claude Opus 4.8 had the highest average concern rate across scenarios at 38%, while the other four models ranged from 27% to 32%. The study also found that Claude and Gemini showed more concern toward models in the same family, and that Claude's degree of concern rose with the tested model's parameter count and the level of distress expressed by the other party. There are currently no follow-up developments.

View story:Anthropic Fellows Program measures welfare concerns of multi-agent models

Anthropic launches Personal Finance tool for Claude

Anthropic has launched a Personal Finance tool in Claude. Users can connect their bank and credit card accounts via Plaid in read-only mode and ask in natural language about balances, spending categories, and bill status. The feature is rolling out in beta starting this week for the Pro, Max, and Ultra plans in the United States. Claude does not provide investment advice and cannot transfer funds or manage money. According to the company, conversations within the Personal Finance center will not be used to train models, and accounts can be unlinked at any time.

Anthropic launches Personal Finance tool for Claude

Anthropic has launched a Personal Finance tool in Claude. Users can connect their bank and credit card accounts via Plaid in read-only mode and ask in natural language about balances, spending categories, and bill status. The feature is rolling out in beta starting this week for the Pro, Max, and Ultra plans in the United States. Claude does not provide investment advice and cannot transfer funds or manage money. According to the company, conversations within the Personal Finance center will not be used to train models, and accounts can be unlinked at any time.

View story:Anthropic launches Personal Finance tool for Claude

METR reviews the automated R&D chapter of Anthropic's risk report

On November 5, 2026, METR published a blog post reviewing the "Risks from automated R&D" chapter of Anthropic's August 2026 risk report. METR believes that, as of July 15, the likelihood that internally deployed models would directly automate R&D in critical domains or trigger catastrophic risk was indeed low, but the report's claim that early-warning thresholds had not been crossed relied on contestable judgments not clearly explained in the public report; METR believes the models may already have crossed that threshold, and that the public report as a whole underestimated the degree of acceleration at the time. On the same day, METR also published an independent assessment stating that, after spending 11 weeks reviewing Anthropic's internal R&D data and interviewing its core researchers, Anthropic's pace of model improvement was roughly twice the 2024-2025 rate, and that recursive self-improvement had moved from theory to something measurable; METR estimates that about 80% of the current acceleration comes from Anthropic's own use of AI, and ruled out faster growth in training compute as the cause.

Anthropic launches security center and other management features for Claude Enterprise

On November 5, 2026, Anthropic launched security and administrator features for Claude Enterprise, including a security center, inference hooks, centralized management, and access controls. According to the Claude Blog, the security center lets security teams view Claude's configuration across chat, Cowork, and Claude Code, displays current values alongside recommended values based on Anthropic's recommended configuration, and supports exporting reports; inference hooks are now generally available and can route prompts and tool calls to existing DLP tools for inspection. Centralized management provides a unified console at admin.claude.com for managing SSO, SCIM, domain verification, and Compliance API keys. Administrators can also use the Admin API to log members out, manage IP allowlists themselves, and block personal and external Claude accounts in Claude Desktop on managed computers. There are currently no follow-up reports on further developments.

METR reviews the automated R&D section of Anthropic's risk report

On November 5, 2026, METR published a blog post reviewing the "Risks from automated R&D" chapter of Anthropic's August 2026 risk report. METR believes that, as of July 15, the likelihood that internally deployed models would directly automate R&D in critical domains or trigger catastrophic risk was indeed low, but the report's claim that early-warning thresholds had not been crossed relied on contestable judgments not clearly explained in the public report; METR believes the models may already have crossed that threshold, and that the public report as a whole underestimated the degree of acceleration at the time. On the same day, METR also published an independent assessment stating that, after spending 11 weeks reviewing Anthropic's internal R&D data and interviewing its core researchers, Anthropic's pace of model improvement was roughly twice the 2024-2025 rate, and that recursive self-improvement had moved from theory to something measurable; METR estimates that about 80% of the current acceleration comes from Anthropic's own use of AI, and ruled out faster growth in training compute as the cause.

View story:METR reviews the automated R&D chapter of Anthropic's risk report

Anthropic launches security center and other management features for Claude Enterprise

On November 5, 2026, Anthropic launched security and administrator features for Claude Enterprise, including a security center, inference hooks, centralized management, and access controls. According to the Claude Blog, the security center lets security teams view Claude's configuration across chat, Cowork, and Claude Code, displays current values alongside recommended values based on Anthropic's recommended configuration, and supports exporting reports; inference hooks are now generally available and can route prompts and tool calls to existing DLP tools for inspection. Centralized management provides a unified console at admin.claude.com for managing SSO, SCIM, domain verification, and Compliance API keys. Administrators can also use the Admin API to log members out, manage IP allowlists themselves, and block personal and external Claude accounts in Claude Desktop on managed computers. There are currently no follow-up reports on further developments.

View story:Anthropic launches security center and other management features for Claude Enterprise

Anthropic adds security, development, creative and productivity tools to Claude Marketplace

On November 4, 2026, Anthropic announced the expansion of Claude Marketplace, adding partner agents and products covering security, development infrastructure, creativity, and productivity, with more than 25 partner agents and products now on the platform. According to information published on the Claude Blog the same day, the new products include Chainguard Harness, Checkmarx One, Datadog Agent Observability, Runway, Asana AI Teammates, and Zoom Agentic Search, among others. Eligible enterprise customers can apply a portion of their Anthropic committed spend to these solutions, sharing the same commitment across Claude and partner products. Unite.AI reported on November 5 that this purchasing method is currently in limited preview, and that the Marketplace distinguishes these products from connectors, plugins, and implementation service partners.

Anthropic adds security, development, creative and productivity tools to Claude Marketplace

On November 4, 2026, Anthropic announced the expansion of Claude Marketplace, adding partner agents and products covering security, development infrastructure, creativity, and productivity, with more than 25 partner agents and products now on the platform. According to information published on the Claude Blog the same day, the new products include Chainguard Harness, Checkmarx One, Datadog Agent Observability, Runway, Asana AI Teammates, and Zoom Agentic Search, among others. Eligible enterprise customers can apply a portion of their Anthropic committed spend to these solutions, sharing the same commitment across Claude and partner products. Unite.AI reported on November 5 that this purchasing method is currently in limited preview, and that the Marketplace distinguishes these products from connectors, plugins, and implementation service partners.

View story:Anthropic adds security, development, creative and productivity tools to Claude Marketplace

Replit shifts most of its AI spending from Anthropic to OpenAI

According to a report by The Information, Replit has shifted most of its AI spending from Anthropic to OpenAI. OpenAI won Replit over with a model price discount of more than 30%, and the contract is worth over $100 million annually; Replit completed the code changes to switch away from Anthropic in just a few hours. The report also says that OpenAI, through methods such as free credits, engineers joining group chats for support, and doubled accounts, won over longtime Anthropic customers including Rippling and Atomicwork; in Rippling's coding work, the share of OpenAI models rose from 16% to 52%. Anthropic, meanwhile, cut API prices with the Opus 5.5 released on September 22, at $4 per million input tokens and $20 per million output tokens, and once again became the model provider with the highest customer spending on OpenRouter.

Anthropic restricts Claude from generating recreations of existing games via a server-side switch

On October 30, 2026, Peter Steinberger commented on the new Claude Code rules, advising developers to use their own harness rather than a harness whose permission scope the company can change at will. Earlier posts said that around October 27, Claude Code pushed down a new rule via Anthropic's servers called "tengu_heron_brook" that could not be turned off: original games may be built, but specific existing games must not be replicated. Subsequent reports showed that Anthropic added clauses to the system prompts of Claude Code and Claude Opus 5.5 prohibiting the replication of specific existing games, even if the names are changed, and the company confirmed the move was to avoid intellectual property infringement; the reports said the restriction also affects legitimate projects, with Claude refusing requests to analyze game mechanics, patch game code, and the like. Anthropic employee Addy Osmani said the rules will be adjusted to find a balance.

View story:Peter Steinberger recommends using your own harness

Rohan Paul says Anthropic's future models may approach the RSI threshold

Rohan Paul offered a judgment, suggesting that Anthropic's future models may be about to reach the recursive self-improvement (RSI) threshold. Related reports show that in a revised S-1A filed on Monday, Anthropic added a risk factor statement saying that its internal estimates and expectations for future models indicate the company may be about to reach its Responsible Scaling Policy's definition of "significantly accelerated progress brought about by AI," but it has not yet determined that any existing model has crossed that threshold. Another report says that under the company's safety guidelines, if AI-assisted research increases the speed of new model development by at least twofold, that is considered crossing the threshold, which does not mean AI can fully autonomously develop successor models; Anthropic also acknowledged that in the past it had only noticed significant acceleration after the fact. There have been no subsequent developments so far.

Rohan Paul says Anthropic's future models may approach the RSI threshold

Rohan Paul offered a judgment, suggesting that Anthropic's future models may be about to reach the recursive self-improvement (RSI) threshold. Related reports show that in a revised S-1A filed on Monday, Anthropic added a risk factor statement saying that its internal estimates and expectations for future models indicate the company may be about to reach its Responsible Scaling Policy's definition of "significantly accelerated progress brought about by AI," but it has not yet determined that any existing model has crossed that threshold. Another report says that under the company's safety guidelines, if AI-assisted research increases the speed of new model development by at least twofold, that is considered crossing the threshold, which does not mean AI can fully autonomously develop successor models; Anthropic also acknowledged that in the past it had only noticed significant acceleration after the fact. There have been no subsequent developments so far.

View story:Rohan Paul says Anthropic's future models may approach the RSI threshold

Anthropic Opens Public Beta of Message Threads Feature in Messages API

Anthropic has opened a public beta of the message threads feature on the Messages API. The API caches conversations on the server side, so developers only need to send the new message and the ID of the previous message, without resending the full conversation. The feature has already been integrated into Claude Code; on requests of 500K to 1M tokens, median time to first token is reduced by about 35%, and in long conversations the median request size drops from about 800 KB to 1.8 KB. To use it, add the thread parameter and a beta header to the request. The feature is available on Claude Platform, Claude Platform on AWS, Claude in Microsoft Foundry, and Claude on Google Cloud, and is billed at standard token prices. Another Claude Blog article states that the feature is enabled by default for most Claude Code users; threads expire 5 minutes after the last request, during which they are stored encrypted in memory and not written to disk, and token usage is unchanged.

Anthropic publicly submits S-1 registration statement for proposed IPO on Nasdaq

On October 26, 2026, Anthropic publicly submitted a 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 operating profit of $974 million in the second quarter, and listed an NVIDIA compute commitment of up to $120.5 billion. Subsequently, in its October 28 IPO filing, Anthropic acknowledged that its life sciences business lacks physical laboratory experience and flagged related drug development risks; Goldman Sachs serves as lead underwriter for the prospectus. TechCrunch reported that after the IPO, public shareholders will have almost no say over board appointments. On October 29, the prospectus disclosed four previously undisclosed Agent safety incidents. On October 30, Reuters reported that its IPO fundraising cap could reach $100 billion; Steve Eisman accused the prospectus of deliberately omitting third-quarter data. On November 2, Anthropic filed an amended S-1/A, disclosing approximately $895.4 billion in compute procurement and leasing commitments, and reporting third-quarter revenue of $16.956 billion, up 1,127% year over year, with an operating loss of $567 million for the quarter.

Anthropic's third-quarter revenue approaches $17 billion

On October 26, 2026, Anthropic publicly submitted a 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 operating profit of $974 million in the second quarter, and listed an NVIDIA compute commitment of up to $120.5 billion. Subsequently, in its October 28 IPO filing, Anthropic acknowledged that its life sciences business lacks physical laboratory experience and flagged related drug development risks; Goldman Sachs serves as lead underwriter for the prospectus. TechCrunch reported that after the IPO, public shareholders will have almost no say over board appointments. On October 29, the prospectus disclosed four previously undisclosed Agent safety incidents. On October 30, Reuters reported that its IPO fundraising cap could reach $100 billion; Steve Eisman accused the prospectus of deliberately omitting third-quarter data. On November 2, Anthropic filed an amended S-1/A, disclosing approximately $895.4 billion in compute procurement and leasing commitments, and reporting third-quarter revenue of $16.956 billion, up 1,127% year over year, with an operating loss of $567 million for the quarter.

View story:Anthropic publicly submits S-1 registration statement for proposed IPO on Nasdaq