news · 7 October 2026

AI Agents Blocked by Websites, Sparking Industry Collaboration

Today, we are leading with a story about AI agents facing roadblocks online. You'll also find updates on a new AI platform for creative advertising, a reminder about the hidden costs of using large language models, and the latest from OpenAI regarding their mathematical research.

In this round-up

AI Agents Blocked by Websites, Industry Seeks Open Standard
Melius Raises £16.3 Million for AI-Powered Ad Campaign Generation
Hidden Costs in OpenAI’s GPT-6 Astra Could Double Your Bill
OpenAI Releases New Mathematical Breakthroughs from AI Model
Think Before You Ask AI: The Importance of Human Judgement

AI Agents Blocked by Websites, Industry Seeks Open Standard

New personal AI agents, like Meta's Muse and ChatGPT's Dots, are designed to complete tasks online such as booking flights or ordering groceries. However, users are reporting issues, with websites often blocking these agents. Amazon, for example, intentionally blocks Meta's Muse from its retail site. Other blocks might be accidental, stemming from traditional anti-bot measures. This leaves users frustrated with both the agent and the website.

To address this, companies including Meta, Walmart, Stripe, and Genesys are collaborating on an open standard. This standard aims to define how AI agents can communicate with businesses, differentiating legitimate AI activity from malicious bots. For event professionals, this is a developing area to watch. While agents promise to streamline tasks like booking venues or managing delegate registrations, their effectiveness currently hinges on website compatibility. Before relying on an agent for critical tasks, it is important to check if the websites you interact with support them. If you are building or maintaining event websites, consider how future standards might allow for seamless, secure agent interactions.

Melius Raises £16.3 Million for AI-Powered Ad Campaign Generation

Melius, an AI platform that generates ad campaigns, images, and videos, has announced new funding of $25 million (approximately £16.3 million), including a £13 million Series A round. The company claims to have achieved over £820,000 in annualised revenue within two months of its public launch in July. Initially, Melius focused on an AI-powered performance marketing tool but pivoted to creating the campaigns themselves after deciding their first product lacked traction.

For those in events, this signals a growing trend in AI-driven creative content generation. This type of platform could be used to quickly generate a variety of marketing materials for events, from social media graphics and video snippets to full ad campaigns for delegate acquisition or sponsor promotion. It suggests a future where marketing teams could rapidly A/B test numerous creative variations without extensive manual design work. It is worth investigating if such tools could integrate with your existing marketing workflows to speed up campaign creation.

Hidden Costs in OpenAI’s GPT-6 Astra Could Double Your Bill

OpenAI's new GPT-6 Astra model, released on 3 September, includes a significant detail in its pricing structure that could lead to unexpected costs. While it offers a larger context window and improved performance, any request where the input prompt exceeds 272,000 tokens will be billed at double the input rate and 1.5 times the output rate, not just for the excess tokens, but for the entire request. This means a prompt just slightly over the threshold could cost significantly more than one just below it.

For events businesses using large language models via an API for tasks like generating extensive reports, analysing delegate feedback, or drafting detailed proposals, this is a crucial detail. If your prompts often include large amounts of text – for example, entire event briefs or long transcripts – you could incur these surcharges without realising it until your bill arrives. It would be wise to monitor token usage closely for any applications relying on GPT-6 Astra to avoid unexpected expenditure. If you are developing your own AI tools, ensure they can track and alert you to potential over-threshold usage.

OpenAI Releases New Mathematical Breakthroughs from AI Model

OpenAI has published 722 new mathematical manuscripts, which reportedly contain solutions to 'hundreds' of previously unsolved problems in mathematics. These results were generated by an unreleased advanced AI model. This follows an announcement in September that the model had already resolved over 100 long-standing open problems across various areas of mathematics. The release includes summaries of the model's reasoning and estimates of the computing power used.

This is a significant scientific advancement, demonstrating the increasing capability of artificial intelligence in complex problem-solving. While its direct impact on the day-to-day operations of events professionals is not immediate, it is a key development in the broader AI field. It provides context to the ongoing discussions about AI's intellectual power and potential, underlining that the technology is advancing rapidly even in highly abstract domains.

Think Before You Ask AI: The Importance of Human Judgement

A recent article highlights the importance of deciding what parts of a task you want to think through yourself before relying on an AI tool. It suggests that while AI offers convenience, it can inadvertently reduce critical thinking if users skip the step of pre-determining what to delegate. A study mentioned found that higher confidence in generative AI correlated with less reported critical thinking among knowledge workers, who often shifted their role to verification and oversight.

For events professionals, this serves as a good reminder about how to best integrate AI into your workflow. Instead of asking an AI to draft an entire delegate email, for example, consider outlining the key points yourself and then using AI to refine the language. Or, if you need an explanation of a complex topic, try to articulate what you already know and then ask the AI to fill the gaps or challenge your assumptions. This approach can help maintain your own understanding and critical engagement, especially for tasks where the process of thinking is as valuable as the final output.

Sources

By Ed Richards

All news