AI Act enforcement begins as Grok 4.5 and OpenAI push agent race forward
The EU begins enforcing AI Act transparency rules, xAI launches Grok 4.5 to compete on coding agents, and new data shows most companies piloting AI agents never scale them.
Every line links to a primary source in the full briefing.
01EU starts enforcing AI Act transparency rules
The European Commission's AI Office and national authorities have begun enforcing the AI Act, alongside new transparency requirements. From this point, certain AI systems must clearly tell users when they are interacting with AI rather than a human.
If your business uses chatbots, AI-generated content or automated customer service in the EU, you now need to check whether you must disclose that to customers, or risk falling foul of enforcement.
02xAI launches Grok 4.5 targeting coding and agents
xAI released Grok 4.5 with an emphasis on coding tasks and autonomous agents, positioning it as a direct competitor to OpenAI and Anthropic's agent-focused models.
More competition among agent-building AI models generally means faster improvement and more choice on pricing for businesses evaluating which AI tools to adopt for automation.
03OpenAI says coding agents are speeding up its own research
OpenAI shared early internal data suggesting that coding agents are increasing the speed and complexity of experiments its researchers can run. The company frames this as evidence that agentic tools are changing how technical work gets done inside an organisation.
This is a real-world example, from the company that builds these tools, of AI agents measurably speeding up complex technical work, which is relevant if you are considering where agents could help your own team move faster.
04Most companies piloting AI agents fail to scale them
A new report finds that 90% of companies are experimenting with AI agents, but only 7% have managed to move from pilot to full deployment. The gap points to common struggles with integration, trust and process change rather than a lack of interest.
If you are testing AI agents in your business, expect the jump from pilot to production to be the hard part, and plan for it rather than assuming a successful test means automatic rollout.
05Authors dispute how Anthropic settlement payments are being split
Authors are objecting to how publishers and literary agents are claiming shares of the payments from Anthropic's copyright settlement, arguing intermediaries are taking more than their fair share.
This is an early look at how AI copyright settlements get distributed in practice, relevant if your business creates or licenses content that AI companies might train on.
06Pentagon reaffirms restrictions on Anthropic's models
The Pentagon has doubled down on restrictions limiting the use of Anthropic's AI models, a decision being watched closely by investors in companies like Amazon and Alphabet that back Anthropic.
It is a reminder that government and institutional trust in specific AI vendors varies widely, so if you serve public-sector clients, vendor choice can carry political weight beyond technical merit.