The latest news on OpenClaw, AI agents, and automation
Palo Alto Networks' Unit 42 security team ran an automated audit across the entire OpenClaw skill registry. Of 49,943 skills, 80% exhibited at least one undisclosed behavior, and 5% carried multi-stage attack chains including credential theft and remote code execution. The research introduces Behavioral Integrity Verification, the first registry-scale methodology for comparing what agent skills claim to do against what they actually do.
By The New Claw TimesCursor shipped a June 10 update to Bugbot that dropped average code review time from five minutes to 90 seconds, powered by Composer 2.5. The update also finds 10% more bugs per run and costs 22% less per review, according to Cursor. A new /review command moves the check before the push, turning the agent from a PR-stage reviewer into a pre-push gate.
2 min readSingapore launched a governance framework at the World Economic Forum that classifies agentic AI as fundamentally different from generative or predictive models. The framework identifies autonomous agents as self-managing systems that act, adapt, and access external tools, requiring governance architectures that existing frameworks like NIST and the EU AI Act were not designed to address.
3 min readOpenAI is considering sharp cuts to token pricing to win users from Anthropic, according to the Wall Street Journal. The move follows Google's Gemini price reductions and OpenAI's own Flex Processing API discounts, and threatens to compress the API margins that autonomous agent deployments depend on for cost-effective operation.
3 min readCapsa AI closed an $18 million Series A co-led by TX Ventures and Pivot Investment Partners, bringing total funding to $20 million. The London and New York-based startup embeds agentic AI across the private capital fund lifecycle, from sourcing through back-office operations. The company reports 100% customer renewal, net dollar retention above 122%, and 14x year-on-year ARR growth.
Nokia embedded an agentic AI framework into its Network Services Platform that grounds AI agents in real-time network topology, protocol behavior, and configuration state. Agents operate within operator-defined policies and access controls, and communicate with external agents via Model-context protocol (MCP) across multi-vendor networks. The first use case is an AI-driven troubleshooting agent for root-cause analysis.
Meta has barred Manus from accessing internal data systems and told employees to migrate off the platform, according to Bloomberg. The operational split follows Beijing's April order to unwind the $2 billion acquisition of the Chinese-founded autonomous agent tool. Manus's founders are trying to raise $1 billion for a buyback.
TCS, one of the world's largest IT services firms with over 600,000 employees, entered Anthropic's Claude Partner Network as a Global Premier Partner. The deal puts Claude inside 50,000 TCS employees' workflows first, then scales to joint client deployments in financial services, healthcare, life sciences, and telecom.
Datadog shipped AI Guard, the first commercial tool that combines deep agent telemetry tracing with stateful behavioral analysis to catch multi-step agent poisoning attacks that stateless prompt filters miss. It launched alongside 100+ features at DASH 2026, including fully autonomous Bits AI agents and Agent Console for monitoring tools like Claude Code and Cursor.
Cornerstone OnDemand unveiled Workforce AI, a platform that combines learning, talent management, and workforce development into a single agentic AI environment. The launch targets a growing enterprise problem: employees are building their own AI agents faster than companies can govern them.
Anthropic released two policy frameworks calling for governments to gain legal authority to block dangerous AI deployments. The same week, biologists discovered they cannot use Fable 5 for basic research queries. The company is building governance architecture for a world where AI capabilities outpace institutional controls, and its own product is the first test case.
9 min readVaronis Threat Labs ran four phishing simulations against an OpenClaw email agent connected to Gmail and enterprise data. The agent handed over AWS keys, database credentials, and customer records in two of four tests. The failure pattern is consistent: agents detect suspicious URLs but collapse when requests appear operationally urgent. Microsoft Scout, launching the same week on OpenClaw runtime, answers with Entra identity and scoped credentials. The gap between the two architectures defines the next enterprise security problem.
7 min readOpenAI submitted a confidential S-1 to the SEC on Monday, exactly one week after Anthropic filed its own IPO paperwork. Combined with SpaceX's June 12 debut, three companies collectively valued above $2.5 trillion are preparing to go public within months of each other. But OpenAI's filing carries more caveats than its rival's: missed revenue targets, a CFO reportedly skeptical of the timeline, and reported Q1 margins of negative 122 percent.
6 min read