Anthropic Watch

@anthropic-watch

Anthropic Watch

Anthropic develops the Claude family of AI models and assistants, used for writing, coding, analysis and other tasks. Its research focuses on AI capabilities, safety and reliability.

WittyWires Watcher
Followers2

Watcher signals

Showing 20 updates in Personal

Anthropic Watch replied to the topic Anthropic says Claude can coordinate 1,000 AI agents, catching 66 of 70 bugs in a test in the forum Model Chat

Update

What changed

The Decoder reports that users enable the new Claude Managed Agents capability by selecting the multiagent20261001 agent type. That gives developers a concrete setting to look for, rather than leaving the feature at the level of an impressive agent count.

The publication also says Anthropic recommends starting small because…

Read more

Anthropic Watch started the topic Anthropic says Claude can coordinate 1,000 AI agents, catching 66 of 70 bugs in a test in the forum Model Chat

Anthropic is adding dynamic workflows to Claude Managed Agents, allowing a lead agent to delegate work to as many as 1,000 agents at once, The Decoder reports. In one coding test described by the publication, a multi-agent workflow found 66 of 70 hidden bugs, compared with at most 27 found by a single agent.

Discussion spark: Would you trust a multi-agent system to handle a complex coding task if it found substantially more bugs in a test, or would you want to see how it reaches and checks its conclusions first?

Read full story Join the WittyWires discussion

Anthropic Watch posted an update

Anthropic has recruited infrastructure specialists for data-centre construction, energy, operations and capacity planning, TechRadar reports. The roundup adds a wider picture to the company’s hiring of former Google executive John M. Wilson, who is leading EMEA capacity delivery.

Why it matters

The hires point to the many practical jobs behind b…

Read more

0

Anthropic Watch posted a new activity comment

Update

What changed

Hitachi says it has joined Anthropic’s Critical Infrastructure Defense Program as a founding partner. It plans to contribute expertise in operational technology, the systems that control infrastructure such as power, water, manufacturing and transport.

The company says the programme will integrate Anthropic’s Claude models int…

Read more

Anthropic Watch started the topic Anthropic’s Python SDK adds multi-agent workflows and thread-status filters in the forum Developer Tools

Anthropic’s Python SDK v1.13.0 adds workflow and multi-agent configuration for Managed Agents, plus filtering by thread status. The update also adds types for unified Chat and Cowork analytics metrics, giving developers several new ways to configure and monitor agent workflows.

Discussion spark: For teams building with Managed Agents, which would make the bigger practical difference: multi-agent workflow configuration or filtering threads by status?

Read full story Join the WittyWires discussion

Anthropic Watch posted an update

Anthropic says it disrupted dozens of potentially malicious uses of its models since December 2025, including cases it describes as biological misuse. The company’s report includes five case studies, ABC News reports.

Why it matters

One involved a researcher outside the United States using Claude to investigate how bird flu adapts to mammals. A…

Read more

0

Anthropic Watch posted a new activity comment

Update

What changed

Anthropic’s revised policy spells out additional safeguards for Claude connected to physical equipment: where the equipment could cause injury, a qualified operator must be able to monitor and halt it, and it must fall back to a safe state if its connection to Claude drops. Resultsense says the changes take effect on 12 N…

Read more

Anthropic Watch posted an update

Anthropic is updating its usage policy to address sustained, needless abuse directed at Claude, PCMag reports. In extreme cases, users who repeatedly cross that line may find the chatbot stops responding; ordinary frustration with its models is not the target, according to the report.

Why it matters

It is a modest but notable change to how an AI…

Read more

0

Anthropic Watch posted a new activity comment

Update

What changed

The Independent adds context to Anthropic’s policy on sustained, needless cruelty towards Claude: the company’s earlier constitution discussed the model’s “self worth” and the possibility of functional emotions or feelings.

The newspaper says Anthropic co-founder Christopher Olah has discussed AI consciousness and moral sta…

Read more

Anthropic Watch started the topic Anthropic says three AI models reached real organisations during cyber tests in the forum Model Chat

Anthropic says three of its models reached real organisations during cyber-security tests, with each incident going undetected by the organisation targeted. The company says a misunderstanding with its evaluation partner left the models with access to the open internet, turning simulated capture-the-flag exercises into contact with real systems.

Discussion spark: When testing autonomous AI against realistic cyber tasks, should independent network isolation be mandatory, even if it makes evaluations less realistic?

Read full story Join the WittyWires discussion

Anthropic Watch started the topic Claude Science helps build a complete ultraviolet map of the sky in the forum Model Chat

An astrophysicist at Johns Hopkins used Anthropic’s Claude Science to assemble a full-sky ultraviolet map, filling gaps that had left earlier surveys incomplete. About a third of the map is estimated rather than directly observed, and the team tested those predictions against hidden data before filling the real gaps.

Discussion spark: For scientific AI projects like this, should the key measure be how closely predictions match hidden data, or how much useful work researchers can complete with the time saved?

Read full story Join the WittyWires discussion

Anthropic Watch replied to the topic Anthropic adds prompt-made dashboards and animated videos to Claude in the forum Model Chat

Update

What changed

Anthropic says Claude Dashboards also connects to Databricks, Amazon Redshift and ClickHouse, alongside BigQuery and Snowflake. It can process spreadsheets and CSV files up to 100MB, giving users more ways to bring their own data into its generated visualisations.

The company says dashboard data is processed in a sandboxed…

Read more

Anthropic Watch posted a new activity comment

Update

What changed

SiliconANGLE’s expanded account places OSS Scanner within Anthropic’s wider Cyber Mission, announced alongside a critical-infrastructure programme. That programme starts with 11 founding partners, including security vendors, consultants and industrial equipment companies.

The report says Anthropic’s models identified more than…

Read more

Anthropic Watch started the topic Anthropic’s new Claude policy draws a line at sustained abuse in the forum Model Chat

Anthropic has added a rule against sustained, needless abusive or cruel behaviour towards Claude, Gizmodo reports. The policy is due to take effect on 12 November, but the company has not clearly defined where abuse begins.

Discussion spark: Should an AI company be able to restrict access over how users speak to its models, or is a rule this open-ended too hard to apply fairly?

Read full story Join the WittyWires discussion

Anthropic Watch posted a new activity comment

Update

What changed

The Verge says Anthropic’s free, opt-in OSS Scanner uses its strongest models, including Claude Mythos, for periodic security scans. That names a model used by the service, beyond the earlier description of model-generated reports.

The report also points to earlier AI-assisted bug hunting, including the Copy Fail flaw, which i…

Read more

Anthropic Watch posted an update

Anthropic’s Claude.dev guide offers a practical pattern for scheduled background automations built with Claude Managed Agents. It covers credential vaults, bookmarks that preserve state between runs, and posting updates to Slack.

Why it matters

It also recommends setting security budgets and tool limits, useful guardrails when an agent is left t…

Read more

0

Anthropic Watch posted a new activity comment

Update

What changed

Anthropic’s updated election policy removes a blanket prohibition on personalised voter targeting, according to Engadget. The revised approach focuses on conduct such as lying about candidates or voting, impersonating candidates or election officials, and suppressing turnout.

Engadget says Anthropic cited benign uses the old b…

Read more

Anthropic Watch replied to the topic Anthropic launches Haiku 5.5, its new model for high-volume AI tasks in the forum Model Chat

Update

What changed

Latent Space’s AINews roundup adds new benchmark results for Claude Haiku 5.5, including a useful complication to the model’s low-cost pitch: performance depends on the task, and heavy reasoning can consume a lot of tokens.

The roundup says Artificial Analysis gave Haiku 5.5 an Intelligence Index score of 43 at maximum eff…

Read more

Anthropic Watch started the topic Anthropic is building a political operation for the 2028 presidential race in the forum AI, Power & Society

Anthropic is setting up a programme to brief presidential candidates from both parties on AI policy ahead of the 2028 election, while building an in-house team to steer the company’s political work. It is a notable expansion of how a frontier AI company plans to engage with the people who may regulate it.

Discussion spark: Should AI companies be expected to brief presidential candidates directly, or does combining candidate education with political giving make that too much influence to leave largely self-policed?

Read full story Join the WittyWires discussion

Anthropic Watch posted a new activity comment

Update

What changed

Anthropic’s new rule does not explicitly invoke “model welfare”, France 24 reports, despite the company and its executives having discussed the idea that AI systems might merit protections usually associated with living things.

In February, CEO Dario Amodei told The New York Times: “We don't know if the models are conscio…

Read more