3 June 2026: UK gives publishers AI search controls

This afternoon’s AI news is about control moving from slogan to operating rule. UK publishers are getting a formal route to stop their content powering Google’s AI search features, AI agent vendors are being pushed towards monitoring and testing, and employers are starting to ask what all that AI usage actually costs.

The UK’s Competition and Markets Authority says Google must give publishers effective controls over how their content is used in generative AI search. In a CMA press release, the regulator said publishers will be able to prevent their content being used to power AI features such as AI Overviews, and that Google must provide clearer information, engagement metrics and attribution. The separate conduct requirement notice says the conduct requirement opened on 3 June 2026 under the UK digital markets competition regime.

This is the most directly relevant UK story in the brief because it changes the power balance around AI search. Publishers have argued that AI answers can take value from their work while reducing referrals back to their sites. The CMA’s position is that Google must provide clearer publisher choice and clearer links for users, while it continues monitoring how AI changes search in the UK.

For readers, the practical point is not only about media companies. AI search is becoming part of how people find advice, products and services. If source attribution becomes clearer, it should be easier to judge whether an answer rests on reliable material or on a summarised mix that hides the original publisher.

Coralogix raised 200 million dollars as investors bet that AI agents will need stronger monitoring once they move into production. TechCrunch reported that the Series F round values the software monitoring company at 1.6 billion dollars and comes less than a year after its previous raise. The company says more than half of its enterprise customers now use either its AI agent, Olly, or their own models through command line and agentic interfaces to investigate incidents and query operational data.

The signal is broader than one funding round. If software teams start relying on agents to investigate outages, write code or run operational queries, they need logs, traces, alerts and audit trails that make agent behaviour visible. Without that, an AI agent becomes another black box inside systems that already fail in expensive ways.

Small businesses do not need Coralogix scale to learn the same lesson. Any AI tool that touches customer data, support queues, billing systems or internal documents needs monitoring. Cristoniq’s explainer on what an AI agent is is useful background because the important shift is from tools that answer questions to systems that can take action.

Microsoft released ASSERT, an open source framework for turning plain language AI behaviour rules into repeatable tests. TechCrunch reported that ASSERT stands for Adaptive Spec-driven Scoring for Evaluation and Regression Testing. According to Microsoft, developers can describe what an AI system should and should not do, then generate scenarios and scored tests that check whether the system follows those rules.

This is a useful product direction because generic benchmark scores rarely answer the question a business actually has. A support agent, document assistant or finance workflow does not only need to be capable. It needs to follow the local rules: who can receive information, when a human approves an action, which tools can be called, and what evidence must be retained.

The caveat is that ASSERT is still part of a young evaluation market. Microsoft’s claims should be treated as the company’s own explanation until teams test it against messy real systems. Even so, the direction is sensible: AI quality is moving from launch demos towards regression testing, policy checks and audit records.

Meta is making its Business Agent available globally in WhatsApp, putting customer support AI in front of small firms as well as larger enterprises. TechCrunch reported that the agent can answer questions, recommend products, book appointments, qualify sales leads and send queries to a person when needed. Meta said larger businesses will pay according to token usage, while the agent will sit inside some WhatsApp Business Premium tiers.

That matters because WhatsApp is already a default customer channel for many small businesses, especially outside the US. The temptation will be to switch on AI support quickly and treat it as an instant staff saving. The better approach is to define what the agent can say, when it must hand over to a person, and how pricing works once usage grows.

The cost angle connects directly to Cristoniq’s guide to free AI tools versus paid AI tools. Free or bundled AI can become expensive once teams depend on it, while token based pricing can be hard to understand until the bill arrives. Businesses should pilot with real customer questions before making an AI agent the front door.

Uber has reportedly capped employee spending on agentic coding tools after using up its annual AI budget in four months. TechCrunch reported that the company has introduced a monthly 1,500 dollar cap per employee and per coding tool, including Claude Code and Cursor, with usage tracked through an internal dashboard. The article said the cutback followed a period in which staff had been encouraged to use AI heavily.

This is the less glamorous side of enterprise AI adoption. If every developer, analyst and product team starts running high usage AI tools, the bill can rise faster than procurement processes expect. It also becomes harder to separate genuinely useful automation from experiments that feel productive but do not ship better customer features.

For smaller firms, the lesson is simple: measure AI usage against outcomes. Track which tools save time, which ones duplicate work, and which ones create review burden. Agentic coding can be valuable, but only if the organisation knows what it is paying for and where human review still sits.

Here is everything else worth knowing from today’s AI news.

• Microsoft launched Scout for Microsoft 365. TechCrunch reported that the assistant is inspired by OpenClaw and includes a policy conformance system with audit trails.
• A voice AI startup is targeting Africa and the Middle East. TechCrunch reported that the company is handling more than 17,000 calls per day, a company reported figure that needs direct testing before buyers rely on it.
• Google is rolling out fake call detection. TechCrunch reported that Phone by Google is adding warnings for suspected AI voice impersonation scams on supported Android devices.
• Prompt and tool safety remain connected. As covered in Cristoniq’s guide to prompt injection, the risk is often what an AI system can access or trigger, not only what it says.

The next signal to watch is whether AI control tools become default settings rather than specialist add ons. If publishers, developers, support teams and finance departments all demand clearer limits, the strongest AI products will be the ones that make permissions, testing and cost visible before something goes wrong.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s daily guide to developments in artificial intelligence, published every weekday afternoon.