Hosts Ashley Coffey and Daniel Hill cover two big AI stories this week. First, Anthropic announced that Claude will now embed invisible watermarks in AI-generated text and attach signed "provenance" metadata (a verifiable digital record of where a file came from) to generated files like images, in order to comply with the EU AI Act. Ashley draws a parallel to the EU's USB-C charger mandate to show how EU tech regulations tend to ripple out and affect US companies and consumers too. Second, they dig into Google Gemini's milestone of surpassing 1 billion monthly active users, with a standout stat: 63% of Gemini users talk to it by voice rather than typing — which may explain why Apple chose Gemini to power its revamped Siri.
Hosts Ashley Coffey and Daniel Hill cover two big AI stories this week. First, Anthropic announced that Claude will now embed invisible watermarks in AI-generated text and attach signed "provenance" metadata (a verifiable digital record of where a file came from) to generated files like images, in order to comply with the EU AI Act. Ashley draws a parallel to the EU's USB-C charger mandate to show how EU tech regulations tend to ripple out and affect US companies and consumers too. Second, they dig into Google Gemini's milestone of surpassing 1 billion monthly active users, with a standout stat: 63% of Gemini users talk to it by voice rather than typing — which may explain why Apple chose Gemini to power its revamped Siri.
Links:
Anthropic now embeds invisible watermarks in Claude’s text and signed metadata in its files to comply with the EU AI Act. (Claude.com)
Google’s Gemini app surges to 1 billion users (TechCrunch)
Connect with Ashley:
Connect with Daniel:
Leave a Review on Apple Podcasts!
Stay curious, stay informed, and stay hydrated! See you next week with more pivotal AI insights.
Ashley Coffey (00:01.186)
Hello and welcome back to AI news that you should know about for August 14th. I'm your host, Ashley Coffey
Daniel Hill (00:08.116)
And I'm Daniel Hill. There's a lot of AI news out in the world. We digest it and give you the weekly highlights as fast as humanly possible. This week on the show, we'll be talking about anthropic embedding invisible watermarks in Claude's text. We'll also talk about Google's Gemini app surpassing 1 billion users. We'll be covering all of that on our show today, but first, here's a quick word from our sponsors.
Ashley Coffey (00:34.754)
Welcome back. Let's dive in. Alright. Anthropic now embeds invisible watermarks in Claude's text and signed metadata in its files to comply with the EU AI Act. This is coming from TechCrunch.
This is coming from Claude.com. So it's interesting to me how much EU policy impacts the United States. And I think about the USB C case as a specific example. And let me explain before I lose you here. The European Union passed the Common Charger Directive in 2022, which officially mandates the USB C as a universal wired charging port for a wide range of electronic devices sold in the EU. So this
Rolled out December 28th, 2024, and applies to mobile phones, tablets, cameras, headphones, headsets, handheld video game consoles, portable speakers, e-readers, basically everything, except for laptops. The laptop rollout is coming out October or April 28th, 2026. So that's still impending and still rolling out. The reason I mentioned that is because EU policy does impact us, and that's a great example of one. Here's the news. Anthropic
Has signed the EU AIs Act Article 502, Code of Practice on Transparency of AI generated content as a provider of both generative AI models and generative AI systems. What the marking commitments mean for Claude. New models will mark AI-generated content from day one. Claude models launched in the EU on or after August 2nd, 2026, will support machine-readable marking at launch.
Generated text will carry embedded watermarks, and generated files will include digitally signed provenance metadata where supported. Making marking works everywhere you use Claude. Marks will apply to output from supported Claude models across Claude Platform API, Claude Code, Claud Cowork, and Claude Tag, and literally everywhere where Claude is embedded or offered worldwide. Some platforms or features may not support marking types yet.
Ashley Coffey (02:45.655)
But Claude is going to help you detect its own marks. Anthropic will support users and other third parties to detect its markings. And the code requires that Anthropic will share information and documentation soon. But existing models are in progress, and the law includes a transition period for anthropic models launched before August 2nd, 2026. And they're working on adding marking support for those models as well. So, how does Claude mark content?
Claude uses two complementary techniques to mark content generated and processed by Claude. First is the watermarks embedded in the text that is, that'll come over in copy paste, and then two: watermarks that are embedded in a digital signature, similar to the digital signatures that you see in like a Gemini image. This is a lot, and when Claude generates a supported file type like a SVG or PNG, it's gonna attach assigned provenance metadata.
And that metadata follows the coalition of content provenance provenance and authenticity. The coalition for content provenance and authenticity open standard, which is used across the industry to record information about content provenance. Daniel, what are your thoughts?
Daniel Hill (03:58.793)
I think this is so fascinating. You and I were talking before the call a little bit about what they were doing from a technical perspective in order to make this work. And you had a great thought that perhaps it's embedded in the font or something along those lines, which I think is quite probable or quite likely. Obviously, there's always going to be some workarounds around that, what those end up being or how it works, I have no idea. But to the way the way you started this,
Explanation, which I think was fantastic, using USB C as an example of what the EU government has required of companies there, requiring this level of security, embedded watermarks to know what is AI text, what is not. I think that is such a huge step for the EU as a whole. Obviously, it will spread to the United States in some way, not from a lawmaker standpoint, but just an availability standpoint. But I think this is a big signal, and we'll see other companies making similar steps.
to be able to detect what was text generated by their AI models so that people can know and reference that and the companies themselves can say, yes, that was generated using our tools or it wasn't.
Ashley Coffey (05:08.723)
Yes. And I'm going to put money on this right now. I bet you six months from now we're going to see a huge explosion of human typers on Upwork and Fiverr. Of people that literally all their job is to human type something.
Daniel Hill (05:20.966)
Right, right.
Daniel Hill (05:25.576)
Yes, I agree. And there may be some other workarounds for that where maybe you use voice dictation or something. But yes, I a hundred percent agree. There will be analog loophole workarounds that people will pay for in order to not n not have it known that they used AI generated text a hundred percent.
Ashley Coffey (05:43.051)
Yep, yep. We're gonna take a quick break, but when we come back, we'll be talking about Google's Gemini app surges to one billion users.
Daniel Hill (05:54.067)
Welcome back. Continuing with our stories, Google's Gemini app has now surpassed 1 billion users. By way of background, Google has integrated Gemini really well into all its products, and it's clear that strategy has worked. They're continuing to add users. The company has been steadily integrating Gemini across its products, search, Google Workspace, Android, and also the standalone Gemini app.
The AI mode that Google has in search has had over 1 billion monthly active users globally. By way of contrast, OpenAI's ChatGPT hit 1 million monthly active users back in June. Now, the CEO of Google has announced that the Gemini app has surpassed 1 billion monthly active users. The CEO of Google also noted it was the 14th Google product to hit the 1 billion mark.
We're just talking today, though, about the Gemini app, not all the other AI channels that Google offers through its variety of products. Google shared numbers about how people are actually using Gemini, with 63% of Gemini users talking directly to the assistant using the voice feature. And I think that's a really important note because although it is great to type, and there are some situations where maybe other people are around and you want to not make it obvious what you're talking with the AI about.
There are times where you feel comfortable to use the voice mode, and I think that is a fantastic feature. He also said that Gemini generates more than 150 million images every day. This is also a thing on iOS as well. Google said that it now has 100 million active users on iOS. Ashley, what are your thoughts on this?
Ashley Coffey (07:39.651)
you picked this one because of that specific line that you mentioned.
The 63% of Gemini users talking directly to the assistant using the voice feature, that's huge because I know for a fact that percentage is not the same across Chat GPT or Anthropic. And I'm curious if this is the same reason that Apple decided to use Gemini as their model to underpin Siri, because people hate talking to Siri so much. And I'm curious if this is the solution to try to fix that because people have such a pleasant experience talking to the Gemini model versus others.
So it'll be interesting to see the trajectory of this, but I'm glad you pulled this article. It's really cool.
Daniel Hill (08:18.612)
I'm not surprised that Gemini has had this level of impact. And as someone who is an Android auto user, to be able to chat with Gemini while driving when I think of an idea or I have a very obscure question that I'd like to have answered, and then be able to sit down on my computer later, look at my Gemini history and see that conversation. It reminds me of an action I want to take or something else I want to research. Gemini is really, really powerful and not surprised at all at how many people are using the voice assistant to your
Ashley Coffey (08:47.968)
Also, I think they have a leg up here because if they had get pe if they have people using the voice assistant already in the ecosystem, then they already have a leg up on whatever wearable they come out with next, whether that be glasses or whatnot, because your AI assistant is only as good as the ecosystem that it has access to. Same thing with like when I put on my Meta Ray Bans I only can talk to Kristen Bell, my Meta person. And she only has access to the Meta stuff, but you know, Gemini has access to my Google and yeah, so on and so forth. So really cool that you covered this.
Daniel Hill (09:04.34)
Great point.
Daniel Hill (09:18.1)
Thanks. That is it for our AI news you should know about this week. The show notes have links to everything we've mentioned on the show. If you want more of this, that's what social media is for. You can find me at Daniel Hill Media on Instagram, threads, and YouTube.
Ashley Coffey (09:32.224)
And I'm Ashley R Coffey89 on Instagram. Want the highlights without the twice-a-week commitment. My newsletter is monthly and my link is in my bio.
Daniel Hill (09:42.344)
Follow the show, find send us a topic that you want covered, and we will see you next week for more AI news that you should know about. Thanks for listening.