Research Period: 10 August to 17 August 2026
Please check the infographic below.
Please review the slide version of the newsletter here
1. Everything Claude writes now carries an invisible signature
What changed
Since 2 August, Anthropic has been embedding an invisible watermark in all text Claude produces. It’s live everywhere Claude runs: Claude.ai, Claude Code, the API, AWS Bedrock, Google Cloud and Microsoft Azure. Nothing was announced to users at the time, so if you’ve used Claude this month, your output is already marked. The trigger is EU AI Act Article 50, but Anthropic has rolled it out globally, so UK members are covered too.
Why it matters
The watermark survives copy and paste, and even light editing may leave the trace intact. Think of it as invisible ink in the spacing and word choices rather than a tag you can strip out. Two honest caveats: the detection API isn’t live yet, so nobody can check anything today, and a positive result only shows content “may have been processed by Claude”. That’s a long way from proof you didn’t write it.
How to use it
- Check client contracts and university policies for wording about AI-assisted or AI-generated text
- Decide now where you’re comfortable using Claude to polish your own writing
- Rewrite in your own words where authorship genuinely matters, rather than lightly editing
- Brief your team so nobody is surprised when detection tools arrive
- Treat any future detection result as a conversation starter, not a verdict
Source: Claude Help Centre | TechCrunch | Axios
2. Gemini 3.7 Flash halves the price and tops the coding benchmark
What changed
Google released Gemini 3.7 Flash on 13 August at introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens, running through December 2026. That’s roughly half what Gemini 3.6 Flash costs. It also topped the FrontierCode 1.1 Main benchmark at 43.6%, ahead of Claude Sonnet 5 and GPT-5.6 Terra, with a one million token context window and 64K output tokens.
Why it matters
A cheaper model that also scores higher is unusual, and it’s arrived within a week of Grok 4.6 cutting prices too. If you run anything through the Gemini API, switching is a model name change, not a rebuild. One geographic point worth stating plainly: Gemini Spark, the consumer agent feature, is not available in the UK, EEA, Nigeria or Switzerland. The API and AI Studio routes most members will actually use carry no such restriction.
How to use it
- Swap the model name in existing Gemini API calls and compare output side by side
- Test it in AI Studio first, which needs no billing setup
- Recalculate the cost of any workflow you shelved because token spend looked too high
- Feed whole document sets into the one million token window rather than chunking them
- Diarise a pricing review for December, when introductory rates end
Source: Google Blog | SiliconANGLE
3. Grok 4.6 cuts API prices and lands in Cursor on day one
What changed
xAI released Grok 4.6 on 12 August at $2 per million input tokens and $6 per million output tokens, with a 500K context window. Cursor added it to its model selector the same day it launched, so if you already use Cursor you can try it immediately.
Why it matters
This is the second week running with a significant frontier-model price cut: DeepSeek V4 Flash last week, Grok 4.6 and Gemini 3.7 Flash this week. Grok 4.6 sits at roughly 60% below GPT-5.6 Sol pricing. For business users who run code, drafts or analysis through Cursor regularly, a new competitive model arriving on day one inside your existing tool is worth a direct comparison rather than a note to come back to later.
How to use it
- Switch to Grok 4.6 in Cursor’s model selector and run a real task before deciding whether to keep it
- Recalculate API costs for any workflow currently running on a more expensive model
- Use the 500K context window for document-heavy tasks that currently require chunking
- Compare Grok 4.6 and Gemini 3.7 Flash on your own workload rather than relying on benchmarks alone
- Factor xAI ownership into any decision about data handling for sensitive business content
Source: xAI
Additional Noteworthy
Gemini hits 1 billion monthly active users: Google’s Gemini app passed a billion monthly users, its fastest-growing product ever, with ChatGPT reaching
ChatGPT ads arrive in the UK: OpenAI expanded in-ChatGPT advertising to the UK on 11 August. Users on no-cost and Go tiers now see ads; paid plans stay ad-free. A new UK advertising channel opens for SMEs via openai.com/advertisers. | OpenAI
Microsoft 365 Copilot: Excel and Power BI grounding rolling out now: M365 Copilot subscribers can now analyse Power BI reports in plain English through M365 Chat, alongside the new Copilot Notebooks. Needs an M365 Copilot licence at roughly £25 per user per month. | Microsoft
SpaceX completes $60 billion Cursor acquisition: SpaceX closed its purchase of Cursor on 14 August. The product carries on as normal for now, but watch for pricing or model integration changes under new ownership. | Bloomberg
Gemini hits 1 billion monthly active users: Google’s Gemini app passed a billion monthly users, its fastest-growing product ever, with ChatGPT reaching
Keep exploring
More from the Guild.
AI Moments lands weekly. The Guild’s sessions and articles sit alongside it and go deeper on the things that matter.