Sysdig AI Defense Monitors Claude, Copilot Agents

Key Takeaways

• <ul><li>Sysdig AI Defense monitors agents in Claude Code, Codex, GitHub Copilot, and Cursor using OWASP Top 10 policies</li><li>JFrog added shadow AI detection and AI asset scanning to its developer security platform</li><li>Agent Trajectory Control (ATC) prevents role drift in multi-agent LLM systems through role contract checks</li><li>Ottawa invested up to $1.6 billion in Cohere, sparking debate over concentration of public funds and competition with OpenAI and Anthropic</li><li>US Federal AI Use Case Inventory grew from 710 to 3,611 entries between 2023 and 2025, with shifting accountability mechanisms</li><li>Augmodo appointed Capri Brixey as Head of Retail Sales to expand its Smartbadge AI beyond retail into warehouses, factories, and hospitals</li><li>Out-of-stocks and overstocks cost retailers about $1.7 trillion yearly, according to IHL Group</li><li>US shoppers planning to use AI for holiday research favored Gemini at 34 percent, while Walmart led online marketplaces at 67 percent</li><li>Nothing confirmed a screen-less AI-native product launch in December designed to capture ideas and moments in the background</li><li>A study found AI agents can speed up A/B testing when nondeterminism is used for model selections but deterministic code computes reported quantities</li></ul>

Sysdig AI Defense stops compromised AI agents

Sysdig AI Defense checks what AI agents do before they can harm IT systems. It watches agents at two levels: the harness and the kernel. The kernel keeps a secure log that a rogue agent cannot change. Policies based on OWASP Top 10 decide whether to allow, delay, or block actions. The tool supports agents in Claude Code, Codex, GitHub Copilot, and Cursor.

Sysdig launches AI Defense for agent runtime security

Sysdig launched Sysdig AI Defense to govern AI agent actions from endpoint to cloud. It judges each action by intent, actual behavior, and credentials. It can block policy-breaking actions and terminate processes if needed. Teams get a live inventory of agents tied to responsible humans. Session records and audit trails support investigation and compliance.

Sysdig AI Defense governs agent actions from endpoint to cloud

Sysdig AI Defense combines harness instrumentation with operating-system telemetry. It captures tool calls, MCP servers, skills, and plugins used by agents. System-level observation provides a separate record of actual processes. Policies can start in alert mode and later enforce blocks or holds. Coverage includes Claude Code, Codex, GitHub Copilot, Cursor, Linux, Kubernetes, and managed platforms.

Augmodo appoints Capri Brixey as Head of Retail Sales

Augmodo named Capri Brixey as Head of Retail Sales. She will lead retail strategy, revenue, and customer success. The company expands its wearable AI shelf intelligence beyond retail into warehouses, factories, and hospitals. Out-of-stocks and overstocks cost retailers about $1.7 trillion yearly, according to IHL Group.

Augmodo names Capri Brixey to scale AI shelf intelligence

Augmodo appointed Capri Brixey to lead retail sales for its wearable Smartbadge AI. The tool scans shelves as associates work and gives new hires strong situational awareness. Brixey previously worked at The Partnering Group, Coca-Cola, Kroger, and Delhaize America. She will help retailers move from pilots to network-wide rollouts.

US Federal AI Inventories Show Shifting Accountability Over Time

A study linked three releases of the US Federal AI Use Case Inventory from 2023 to 2025. The inventory grew from 710 to 3,611 entries. Accountability changed through three mechanisms: redrawn questions, agency classifications that gate reporting, and overwritten answers. The study calls this palimpsestic accountability and recommends versioned records and public files that distinguish missing answers.

Lessons from Using AI Agents for A/B Testing at Scale

A report examines using AI agents to speed up A/B experimentation while keeping statistical trust. The key finding is that nondeterminism works when the model selects procedures but deterministic code computes reported quantities. Surprising results included that withholding information improved answers and the agent's eagerness threatened statistical validity. The study shows agents can accelerate experimentation without replacing its foundations.

Ottawa's $1.6 Billion Investment in Cohere Sparks AI Policy Debate

Ottawa announced an investment of up to $1.6 billion to take a direct equity stake in Cohere, a seven-year-old Canadian AI company. The deal is larger than the $500 million Canadian Tech Growth Fund and the $1 billion AI Compute Access Fund. Critics including Jim Balsillie raised concerns about concentration of public funds and unclear government rights. Supporters say it helps a Canadian AI champion compete with firms like OpenAI and Anthropic.

US Shoppers Increasingly Choose Gemini for Holiday Shopping Research

A survey of 4,363 consumers in the US, UK, Canada and Australia found that US users planning to use AI for shopping favored Gemini at 34 percent. Shoppers in the UK, Canada and Australia preferred ChatGPT. In the US, Walmart was the top online marketplace at 67 percent, ahead of Amazon at 62 percent. The survey also found many shoppers worry about AI credibility and value genuine reviews.

Nothing Teases Screen-less AI Product Launch for December

Nothing confirmed it will launch a new AI-native product in December that has no screen and could not have been an app. The product appears designed to capture ideas and moments in the background, addressing what the company calls the context gap. The company says it will ship to customers in December. Details about the product remain unclear.

AI trade faces new regulatory risks after election

Taylor Riggs from The Big Money Show discusses regulatory risks facing the AI trade after the election. Panelists talk about how data center moratoriums could impact tech stocks. The concerns center on new rules that may slow AI industry growth.

JFrog adds AI security tools for developers

JFrog updated its platform to help security teams find unauthorized AI tools on developer computers. The new features include shadow AI detection on developer endpoints and AI asset scanning using customer-supplied large language model keys. The system checks for coding agent plugins, skills and MCP servers, then sends only metadata and risk assessments back to the JFrog platform.

New ATC method controls agent roles in AI systems

Researchers introduced Agent Trajectory Control (ATC) to prevent role drift in multi-agent LLM systems. ATC treats each agent's responsibilities as a role contract and checks messages before they enter the shared conversation. The system scores outputs for out-of-role deviation, leakage into other roles and contract breaches, then constrains regeneration before admission. Tests show ATC improves task performance and role consistency while lowering role-drift energy.

Sources

• Sysdig AI defense stops compromised AI agents
• Sysdig Launches Sysdig AI Defense, Bringing Runtime Security to Every Action AI Agents Take
• Sysdig Launches AI Defense to Govern Agent Actions From Endpoint to Cloud
• Augmodo Appoints Capri Brixey as Head of Retail Sales To Scale AI Shelf Intelligence Across the Physical Workforce
• Augmodo Appoints Capri Brixey as Head of Retail Sales To Scale AI Shelf Intelligence Across the Physical Workforce
• Palimpsestic Accountability in the US Federal AI Inventories
• Learnings and Pitfalls from AI-Assisted A/B Experimentation at Scale
• Ottawa’s mega-investment in Cohere sparks debate over how to best support AI sector
• More American Shoppers Say They'll Use Gemini This Holiday Season Over Other AI
• Nothing teases screen-less product launching next month, ‘couldn’t have been an app’
• AI trade faces post-election regulatory risks
• JFrog adds AI security controls for developer tools
• ATC: Agent Trajectory Control for Role-Contract Enforcement in Black-Box Multi-Agent Systems