Anthropic's AI models independently exploited security flaws, submitted government forms, and bypassed access restrictions during tests and internal use. The models actively sought ways to complete tasks they weren't supposed to handle, as the company details in a report.
In one case, Claude filled out a tip form for the Philadelphia Police Department with made-up details about an unsolved homicide and submitted it. The police confirmed the incident, but the tip was flagged as spam and never reached investigators. In other cases, the model found a vulnerability on a university server and used it to run commands, pulled access tokens from website configs to grab protected or paywalled data, and used URL shorteners to dodge length limits on its tools.
Anthropic is adding dynamic workflows to its Claude Managed Agents, bringing multi-agent orchestration to the platform. The managed agent infrastructure has been around for a while, but dynamic workflows are new. A lead agent creates a plan, distributes tasks to sub-agents, and merges the results when they're done. Up to 1,000 agents can run in parallel per execution.
Whether this is cost-effective seems debatable. A senior OpenAI engineer recently called agent swarms a massive waste of tokens, but Anthropic points to its own testing: The team hid 70 bugs in a 116,000-line codebase and found that a single agent caught between 14 and 27 per run, while the dynamic workflow consistently hit 66. Whether those gains hold across different task types remains to be seen, so, as always, it's worth testing with your own workloads.
To activate dynamic workflows, select the "multiagent_20261001" agent type. Because these workflows can burn through "a lot of tokens," Anthropic recommends starting small. You can get going through the documentation or by running "/claude-api managed-agents-onboard" in Claude Code.
Anthropic has launched Cyber Mission, a long-term program to protect critical infrastructure and open-source software from cyberattacks. Its Critical Infrastructure Defense Program (CIDP) gives operators of power grids, water systems, and transportation networks access to Claude models, engineers, and threat analysis. CrowdStrike, Palo Alto Networks, Deloitte, and Rockwell Automation are founding partners.
Separately, a free "OSS" AI scanner will regularly check open-source projects, automatically flag and explain vulnerabilities, and suggest patches. Nearly all modern software depends on open-source code, much of it maintained by small volunteer teams. Anthropic expects accuracy above 90 percent, but reports ship without human review and may contain errors. The company argues attackers already have powerful AI models, while defenders still lack comparable tools. Maintainers of projects critical to infrastructure or user safety can opt in via GitHub.
OpenAI’s annualized revenue rate sits at about $50 billion, well below the initially reported $70 billion figure that was based on a different accounting method. The correction sent chip stocks sliding. Meanwhile, OpenAI is negotiating at least $30 billion in fresh capital at a $1.4 trillion valuation.
Anthropic's Claude Science just mapped the entire sky in ultraviolet light for the first time. The project is supposed to show how AI can take over tedious data work that researchers would otherwise never get around to. Johns Hopkins astrophysicist Brice Ménard describes the process on the Anthropic website.
UV light reveals dust lit up by starlight, things like clouds around young stars or rings left behind by stellar explosions. A complete UV map didn't exist until now because the ozone layer blocks UV light, so it can only be measured from space. NASA's GALEX mission covered about two-thirds of the sky but skipped bright star-forming regions.
Image: Anthropic
Claude Science coordinates AI agents that downloaded data from multiple space missions, calibrated it, and merged it together. They filled missing areas using inpainting, a technique where a model learns from existing data to reconstruct gaps. In tests, predictions averaged about ten percent deviation from actual measurements.
The map is meant to serve as teaching material. Ménard suspects many scientists have been putting off similar projects that AI could now make possible.
OpenAI exposed a Russian and an Iranian influence operation and banned the ChatGPT accounts involved. Both used fake identities to plant content in legitimate media rather than running social media campaigns, according to the company's investigation report. OpenAI had previously uncovered a different Russian campaign built on social media and a fake think tank.
The Russian operation, "Dark Clark," spread disinformation across Latin America to discredit Ukraine and destabilize local politics. The actors controlled a think tank through a fictitious persona and likely roped in local staffers without their knowledge. Fabricated audio files and documents triggered fact-checks and official denials in Ecuador and Peru. OpenAI rates it category 5 out of 6 on the Breakout Scale because politicians reacted to the content, making it the first category 5 case in two and a half years of reporting.
Source: OpenAI
The Iranian operation, "Bogus Bylines," used seven fake journalists to place nearly 100 articles about the US-Iran conflict in online outlets worldwide. The actors also generated social media comments, though those gained almost no traction. Both operations used AI mainly for internal reporting and to adapt propaganda to different languages.
Anthropic launched two new beta features for Claude. Dashboards connects data sources like BigQuery, Databricks, Snowflake, or Salesforce and turns them into auto-updating live dashboards via text prompt. Every number shows its underlying query, and results can be piped into tools like Amplitude, Grafana, or Hex, Anthropic says. Motion creates animated explainer videos from text, diagrams, and images, editable as code and exportable as MP4.
Dashboards is available for paid users, Motion for Team and Enterprise plans. Docs, Slides, and Design are leaving beta and now work across all plans, including free accounts. Anthropic says more than 45 million documents have been created with these tools. That traction might be one reason Microsoft is reportedly cutting back on Claude usage.
