Anthropic's financial picture came into sharper focus with its S-1 filing in late October and early November 2026. Revenue hit $16.96 billion in Q3, a 47% jump from the prior quarter, but the company posted a $567 million operating loss in that quarter and an $80.7 billion net loss for the first half. Nearly half of first-half revenue flowed through AWS, Google Cloud, and Azure, underscoring the company's heavy reliance on the major cloud providers.
Security around AI agents is tightening from multiple directions. SentinelOne integrated its Singularity Platform directly into Anthropic's Claude Cowork virtual machines on both macOS and Windows, letting security teams monitor agent behavior, receive alerts, and block attacks automatically. Neo Security took a similar approach with a sensor that tracks activity inside Claude Cowork VMs, tying each action back to a specific device, user, and session. Both integrations give customers control over when and how they roll out protections.
NVIDIA released Auto Ontology, an open library under the Apache 2.0 license that helps AI agents write accurate SQL queries against business data. The system uses Nemotron models as defaults and runs locally on Docker Compose or Kubernetes through Helm. NVIDIA reports the tool speeds up ontology creation by about 90 percent.
On the investment front, anxiety about an AI stock bubble is growing. The European Central Bank flagged high AI stock valuations, bond market stress, and shadow bank losses as key financial stability risks, with Vice-President Boris Vujcic set to detail concerns in a biannual review on November 18. A separate survey from a16z found that 87% of CISOs expect cybersecurity spending to rise in 2027 due to AI-enabled threats, and nearly half plan to hire more staff. AI security is now the top priority for security leaders heading into next year.
SAP opened its Max Success Plan to all customers, offering strategic guidance for enterprise AI transformation. This move responds to findings from RAND that 80 percent of AI initiatives fail to deliver value, and MIT research showing that partnering with outside experts doubles deployment success rates. Meanwhile, Dartmouth Provost Santiago Schnell resigned after a committee determined he used AI to revise his writing in a Washington Post opinion piece. Schnell maintained he only used the tool for language refinement. Matthew Delmont will serve as interim provost until June 2028.
Key Takeaways
- <ul><li>Anthropic reported $16.96 billion in Q3 2026 revenue but a $567 million operating loss and $80.7 billion first-half net loss, with nearly half of first-half revenue flowing through AWS, Google Cloud, and Azure</li><li>SentinelOne integrated its Singularity Platform directly into Claude Cowork VMs on macOS and Windows, enabling real-time monitoring and automatic attack blocking for AI agent activity</li><li>Neo Security launched a sensor that traces Claude Cowork actions to specific devices, users, and sessions, backed by Andreessen Horowitz and Bessemer Venture Partners</li><li>NVIDIA released Auto Ontology under Apache 2.0 license, claiming 90 percent faster ontology creation using Nemotron models as defaults</li><li>87% of CISOs expect cybersecurity spending to rise in 2027 due to AI-enabled threats, with AI security now the top priority according to a16z survey</li><li>The European Central Bank identified high AI stock valuations, bond market stress, and shadow bank losses as key financial stability risks</li><li>SAP opened its Max Success Plan to all customers amid research showing 80 percent of AI initiatives fail to deliver value</li><li>Dartmouth Provost Santiago Schnell resigned after using AI to edit a Washington Post opinion piece, with Matthew Delmont serving as interim provost until June 2028</li><li>The White House refused to release documents on its AI security vetting process, citing national security concerns, prompting legal action from Protect Democracy</li><li>EvoMCP improved AI agent performance by 16.5%, 10.2%, and 12.6% on Toolathlon, MCPMark, and MCP-Atlas through evolving training data and tasks</li></ul>
SentinelOne brings security to Anthropic Claude Cowork
SentinelOne integrated its security platform with Anthropic's Claude Cowork. The integration places security inside the virtual machine where Cowork runs. This gives security teams visibility into AI agent activity. The tool supports macOS and Windows devices. Customers control when the integration is turned on.
SentinelOne extends runtime security into Claude Cowork
SentinelOne launched integration with Claude Cowork to cover agentic work inside virtual machines. The Singularity Platform runs inside the Cowork VM on employee devices. Security teams can monitor agent behavior, receive alerts, and stop attacks automatically. The integration supports macOS and Windows. IT teams control rollout through device management.
Neo Security adds visibility inside Claude Cowork VMs
Neo Security released a sensor that monitors activity inside Claude Cowork virtual machines. The tool combines real-time VM visibility with its Neoverse intelligence engine. Security teams can trace each Cowork action back to a device, user, and session. Neo is backed by Andreessen Horowitz and Bessemer Venture Partners. The product is available to all Neo customers.
AI investment bubble worries drive investor caution
Growing concerns about an AI investment bubble are pushing investors to be more careful. More people are watching the market closely as the AI trade faces uncertainty. The worries come as artificial intelligence stocks experience big swings. Investors are searching for ways to protect their money. The trend reflects broader anxiety about AI spending.
3 investment strategies to prepare for a possible AI bubble burst
Fears that the AI stock bubble could burst are returning as the AI market faces turbulence. Business Insider outlined three strategies to help investors prepare. The article offers guidance for protecting portfolios if AI stocks fall. It addresses how investors can adjust their approach during market stress. The advice comes as the AI trade gets rattled.
White House keeps AI security vetting process secret
The White House refused to release documents showing how it reviews AI security threats. It said the information must stay secret to protect national security. Officials found 60 pages of documents but chose not to share them. Ian Bassin from Protect Democracy said the group will keep fighting in court. He said this process is not how regulation should work in the American system of government.
Dartmouth provost resigns after using AI in writing
Dartmouth Provost Santiago Schnell resigned after a committee found he used AI to change his writing. The committee said he violated publication and newspaper policies on AI use. He was accused of using AI in an opinion piece for The Washington Post. Schnell said he only used AI for language refinement and editing. Matthew Delmont will serve as interim provost until June 2028.
EvoMCP improves AI agents through evolving training data
EvoMCP is a method that helps MCP agents improve by adjusting both tasks and training data. It uses a closed loop of task evolution, trajectory evolution, and supervised fine-tuning. The method finds learning targets from student feedback and creates tasks to address weaknesses. Experiments on Toolathlon, MCPMark, and MCP-Atlas showed gains of 16.5 percent, 10.2 percent, and 12.6 percent. This approach helps agents adapt as their skills develop.
NVIDIA releases Auto Ontology for AI agent data queries
NVIDIA released Auto Ontology, an open library under the Apache 2.0 license. It creates a semantic layer that helps AI agents write correct SQL queries against business data. The system uses four core services including a Next.js frontend and FastAPI backend. It runs on Docker Compose locally or Kubernetes through Helm. NVIDIA reports 90 percent faster ontology creation and uses Nemotron models as defaults.
SAP expands transformation guidance for AI customers
SAP made its Max Success Plan available to every customer as AI adoption grows. The plan provides strategic guidance for enterprise transformation. Research from RAND found 80 percent of AI initiatives fail to deliver value. MIT research found that partnering with outside experts doubles deployment success. SAP says consistent guidance helps organizations get real returns from AI investments.
ECB warns of AI stock risks and bond market stress
The European Central Bank identified high AI stock valuations, bond market stress and shadow bank losses as key financial stability risks. ECB Vice-President Boris Vujcic shared these concerns on Monday. The bank will publish its biannual Financial Stability review on November 18.
AI security becomes top priority for CISOs in 2027
A survey by a16z found that AI security is now the top concern for CISOs planning for 2027. About 87% expect cybersecurity spending to rise next year because of AI-enabled threats. Nearly half plan to hire more staff and only 4% plan cuts. Agents using employee permissions are a growing worry because they can slip past existing security tools.
Anthropic IPO filing raises questions about AI spending
Anthropic filed its S-1 on October 26, 2026 and amended it on November 2, 2026. Revenue reached $16.96 billion in Q3 2026, up 47% from the prior quarter. However the company posted a $567 million operating loss in Q3 and a first-half net loss of $80.7 billion. Nearly half of first-half revenue came through AWS, Google Cloud and Azure.
Sources
- SentinelOne Expands AI Security to Anthropic's Claude Cowork
- SentinelOne Expands AI Security Coverage to Claude Cowork
- Neo Brings Security Visibility Inside Claude Cowork
- As concerns over the artificial intelligence (AI) investment bubble grow, more and more people are l..
- How to invest to prepare for AI stock bubble burst: 3 strategies
- White House refuses to disclose how it vets AI security threats
- Dartmouth official resigns over AI use
- EvoMCP: Co-evolving MCP agents and training data
- NVIDIA Releases Auto Ontology: An Open Semantic Layer That Grounds AI Agent SQL
- AI and Transformation Guidance for Every Customer
- ECB flags AI valuations, bond stress and fund losses as main risks
- a16z Survey: AI Security Now Tops Every CISO Priority for 2027
- Anthropic IPO filing tests the sustainability of the AI trade
Comments
Please log in to post a comment.