Weekly Dispatch · archived

Weekly Dispatch · Week 19 of 2026

← latest week

14 items across 4 topics

Crawling & Publisher Controls

This week's discussion highlights the growing inadequacy of traditional robots.txt for AI crawlers, with new proposals like ai.txt and llms.txt emerging for purpose-based content control, and a shift in crawler policy to a strategic, board-level concern for publishers navigating AI search visibility and monetization.

Agents

This week's key discussions on AI agents primarily focus on the evolving landscape of agent security, emphasizing new risks like prompt injection, over-privileged agents, and tool misuse in enterprise deployments, and the need for specialized security architectures beyond traditional controls.

  • witness.ai field-report #AgentSecurity
    Multi agent security: risks and how to secure AI agent systems

    Examines the fundamental shift in security for multi-agent AI systems, detailing new risks and the need for runtime defense and continuous visibility.

    "Multi-agent security is not an extension of single-model security—it is a fundamentally different problem driven by autonomous systems that plan, delegate, and act across environments."
  • witness.ai commentary #AgentSecurity
    AI agent security: how to protect autonomous systems without slowing down the business

    Outlines the biggest AI agent security risks, including prompt injection and excessive permissions, and proposes controls for safe enterprise deployment.

    "The moment an agent can read data, make decisions, and take action inside your business systems, your risk profile changes."
  • witness.ai research #AgentSecurity
    What Is AI Agent Security? Risks & Fixes In 2026

    Explores hidden threats and failures in agentic AI systems, arguing that traditional RBAC is insufficient and a data control layer is essential.

    "Autonomous AI agents pull data from systems you secured, pass it between pipelines, write to memory, and act on instructions without a human reviewing each step."
  • monday.com field-report #AgentSecurity
    Secure AI agents at scale using Microsoft Agent 365 | Microsoft Learn

    Discusses new security challenges introduced by AI agents at scale, such as agent sprawl and tool misuse, and how a distributed security model can address them.

    "AI agents introduce new security challenges, including: Agent sprawl from user-created and SaaS agents that expands the attack surface."

Copyright & Legal

This week saw commentary on artists' role in addressing AI's copyright impacts and a UK government initiative to support smaller creative organizations in licensing their content for AI.

  • artshub.com.au commentary #CopyrightReform
    Artists can help us fight AI's existential threats

    This commentary argues artists are uniquely positioned to address AI's copyright theft and calls for regulation of AI training and updated copyright laws.

    "Artists are disproportionately impacted by AI's tendency to steal work from creatives and deprive them of paid work – but artists are also uniquely qualified to explore the complex issues we now face."
  • qna.files.parliament.uk legal #GovernmentPolicy
    Wednesday 29 April 2026 Daily Report of Written Answers and Written Statements

    A UK Parliament report details a government working group exploring support for independent and smaller creative organizations in licensing their content for AI.

    "In particular, the 18th March 2026 Statement on Copyright and AI Progress announced a working group on independent and smaller creative organisations to explore whether there is a role for government to support their ability to license their content."

Web Ecosystem & AI Impact

This week's reporting highlights the ongoing tension between Google's growing search revenue, attributed to AI Overviews, and independent research indicating significant declines in publisher traffic and organic click-through rates, particularly impacting smaller media outlets.