AI Agents Need Action Monitoring, Says Experts; Passport System Proposed

Key Takeaways

• Security experts say AI agents need action-based monitoring, not just login verification
• A passport and visa system is proposed to limit AI agent authority to specific tasks
• Microsoft earned $10 billion from OpenAI in Q3, with total FY2026 revenue reaching $24.1 billion
• SAP and NTT partner on sovereign cloud and AI services for governments and defense
• WEX and Euclid AI report 80-90% cost reduction per transaction in trucking fleets
• Otto launched the first voice AI CRM for field sales representatives
• Veraify now tracks token spending across approved and unapproved AI tools
• Anthropic's Claude serves 271 million monthly users under an 80-page Constitution
• New OptABA-PC method improves accuracy in causal discovery with formal guarantees
• PHP2Shell attacks highlight risks of centralized data storage in the AI era

AI security must verify actions after login not just users

Security experts say traditional login checks are not enough for AI agents. Unlike regular software, AI agents can make independent decisions after gaining access. Companies must monitor every action an agent takes and match security controls to how independent the agent is.

Passport and visa model proposed for AI agent security

A new article suggests using a passport and visa system to manage AI agent access. The passport would prove an agent is known and trusted. The visa would grant temporary and specific permission for a task. This approach aims to make agent authority limited and tied to a verified human or business process.

Veraify tracks AI token spending across approved and shadow tools

Veraify added token tracking to its AI security platform. The tool monitors token use across approved services and unapproved Shadow AI. It shows spending by department, user, service, and model. This helps companies see total AI costs and improve governance as usage grows beyond pilot projects.

WEX and Euclid AI partner to cut trucking transaction costs

WEX partnered with Euclid AI to bring AI agents to trucking fleets. The AI agents handle paperwork like bills of lading and driver records. Euclid says the technology lets workers handle five to ten times more work while cutting costs per transaction by 80% to 90%. No specific launch date was announced.

Otto launches voice AI CRM for field sales reps

Otto launched what it calls the world's first voice AI CRM. Field sales reps can call the platform and speak to update deals, save contacts, and create tasks. The system uses voice AI models trained on real sales calls and keeps context from emails, calendars, and ERP systems. Reps confirm changes before they are sent.

Three AI cloud stocks investors are watching

Markets are tense as the Federal Reserve considers its next move and companies like Microsoft and Alphabet prepare to share AI spending updates. Everpure, Tower Semiconductor, and Zscaler are three stocks being watched. Everpure provides data storage hardware and cloud management services. It earned US$4.3 billion from computer storage devices. The full screen found 48 more AI and cloud companies to analyze.

AI systems need administrative lawyers

About 271 million people use Anthropic's Claude each month. Claude follows rules set out in an 80-page document called its Constitution. This document sets limits and priorities for the AI model. However, it is not a true constitution and raises questions about how it will be updated and enforced. Experts say AI systems need proper administrative oversight to handle these challenges.

Protecting data from AI era cyberattacks

PHP2Shell attacks using two critical website vulnerabilities forced organizations worldwide offline. AI helped security researchers find these weaknesses from code written in 2012. Cloudflare recorded millions of attempts to exploit the gaps. The article argues that storing data in multiple locations makes it harder to lose. Central data collection creates risks from criminals and foreign powers.

Microsoft earns $10 billion from OpenAI

Microsoft made $10 billion from OpenAI in the quarter ending September 30. For fiscal year 2026, total OpenAI revenue reached $24.1 billion. Microsoft has $7.2 billion in receivables from OpenAI. The companies loosened their exclusive partnership in October 2025. OpenAI agreed to buy $250 billion more in Azure services. Microsoft's ownership of OpenAI dropped from about 25% to 24%.

New method fixes errors in causal discovery

Causal discovery uses tests that can conflict in small samples, leaving no complete causal graph. The new OptABA-PC method formulates repair as a minimum-cost problem using Answer Set Programming. It provides soundness and completeness guarantees with formally guaranteed redress under user contestations. Tests show OptABA-PC retains more correct facts and improves graph reconstruction over Causal ABA on common benchmarks.

SAP and NTT team up for sovereign cloud and AI services

SAP and NTT plan to work together to deliver trusted sovereign cloud and AI capabilities to customers. The collaboration will combine SAP Sovereign Cloud On-Site with NTT's global infrastructure and industry AI expertise. It aims to help governments, defense organizations, and regulated industries modernize critical processes. The companies want customers to benefit from cloud and AI innovation while keeping control over sensitive data, operations, and security.

Sources

• Why 'never trust, always verify - repeatably' security strategy for AI Agents, starts after login
• Passports and Visas: A practical identity model for AI agent security
• Veraify adds token tracking for shadow AI spending
• WEX, Euclid AI target 90% reduction in trucking transaction costs
• Otto Launches World’s First Voice AI CRM, Letting Field Sales Reps Update Deals with a Phone Call
• 3 AI Cloud Stocks Retail Investors Are Watching Before Big Tech Earnings
• AI Constitutions Need Administrative Lawyers
• Securing data in the AI age
• Microsoft recorded $10 billion in revenue from OpenAI
• Optimal Correction Sets for Provable Redress of Contestable Causal Discovery
• SAP and NTT to Cooperate on Delivering Trusted Sovereign Cloud and SAP Business AI Capabilities