AI News This Week: What UK Marketers Need to Know

If you work in marketing and you're trying to keep up with AI news, this week was a lot. Seven announcements landed. Most of them got buried under buzzwords and breathless LinkedIn posts.

Here's what actually happened, ranked by how much it matters to your work. I've skipped the hype and focused on what you can do with each one right now.

Claude's ecosystem expansion changes how agencies work with AI

Anthropic rolled out a marketplace, plugin system, brand voice controls, and scheduled task automation for Claude. This is the biggest shift in how Claude works since Projects launched.

What this means for agencies: you can now set up Claude as a persistent workspace for your business. Your brand voice loaded in. Your client context saved. Tasks running on a schedule without you opening the app.

If you've been treating Claude as a fancy chatbot, this is the update that should change your mind. It's closer to an operating system for knowledge work now.

What to do: Set up a Claude Project for your agency. Add your brand guidelines, client briefs, and standard operating procedures. Start using it as a workspace, not a search bar.

Claude memory on the free plan removes the last excuse

Claude now offers persistent memory, the ability to import your ChatGPT conversation history, 150+ tool connectors, and scheduled tasks. All on the free plan.

This matters because the number one objection I hear from business owners is cost. "I'm not paying for another subscription until I know it works." Fair enough. Now you don't have to.

The memory feature means Claude remembers your preferences, your business context, and your past conversations. You're not starting from scratch every time. And importing from ChatGPT means you don't lose what you've already built.

What to do: Sign up for Claude (free), import your ChatGPT history if you have one, and tell it about your business. Give it a week before you judge it.

ChatGPT for Excel is the one your finance team will care about

OpenAI and Microsoft launched a proper ChatGPT integration inside Excel. Not a plugin that opens a sidebar. An actual integration that manipulates spreadsheets, writes formulas, analyses data, and generates charts.

For agencies that live in spreadsheets for reporting, campaign tracking, or client budgets, this could save hours every week. Ask it to summarise a data set, clean up messy imports, or build a pivot table from natural language.

What to do: If your team uses Excel daily, get someone to test this for 30 minutes on a real task. Not a demo. A real spreadsheet with real data. That's the only way to know if it fits your workflow.

GPT-5.4 and what Sam Altman is saying about it

Sam Altman publicly praised GPT-5.4 as "great at coding, knowledge work, and computer use." It's the latest model from OpenAI, and early reports suggest it's a meaningful step up from GPT-4o for complex tasks.

For agencies, the practical takeaway is this: if you're using ChatGPT for longer, more complex tasks like writing strategy documents, analysing competitor content, or building presentations, GPT-5.4 should produce noticeably better output.

What to do: If you have a ChatGPT Plus subscription, check which model you're using. Switch to GPT-5.4 for your next complex task and compare the output.

Claude Opus 4.6 found 22 Firefox vulnerabilities

Anthropic partnered with Mozilla and put Claude Opus 4.6 to work on the Firefox codebase. It found 22 genuine security vulnerabilities. This isn't a marketing stunt. These were real bugs that human reviewers had missed.

Why this matters beyond tech: it shows that AI is now genuinely useful for quality assurance at a level that used to require specialist human reviewers. The same principle applies to content QA, compliance checking, and brand consistency reviews.

What to do: Nothing immediate. But if you're sceptical about AI handling quality-sensitive work, this is worth reading about.

OpenAI's Codex security agent enters research preview

OpenAI released a security-focused AI agent called Codex that detects code vulnerabilities. It's in research preview, meaning it's not ready for production use yet.

For agencies with in-house dev teams or client websites, this is worth watching. Automated security scanning is becoming more accessible, and AI-powered vulnerability detection is likely to become standard within a year.

What to do: If you build websites or apps for clients, bookmark this. When it moves out of preview, it could replace or supplement your current security scanning tools.

Claude Code remote control for enterprise teams

Claude Code v2.1.69 introduced remote AI coding capabilities for Team and Enterprise plan users. Admin controls let managers set permissions and monitor usage.

This is primarily relevant to larger agencies with development teams. If you're a solo founder or small team, this won't change your day-to-day yet. But it signals where Claude Code is heading: from a solo tool to a team platform.

What to do: If you manage a dev team, explore the Team plan pricing and see whether centralised Claude Code access makes sense for your workflow.

What to focus on this week

If you only act on one thing from this list, make it the Claude ecosystem expansion. Setting up a Claude Project with your brand voice and client context is a 30-minute job that pays off every single day after.