Welcome to Prompt & Circumstance – your weekly deep dive into the evolving world of Generative AI and its impact on marketing and communications.   

Like this newsletter? Share this with a friend or colleague! 

Cannes, Chips, Courts, and Carbon 

Last week, Gen AI moved from potential to execution. At Cannes, the industry’s biggest creative players showed off real-world AI results, not just prototypes. There was a noticeable change in tone from agency execs who are increasingly embracing the technology. Meanwhile, big Tech is racing to build custom chips to stay competitive, even as Nvidia’s dominance grows; Google is rethinking how AI shows up in search; and  the courts are starting to define the legal boundaries of Gen AI and copyright. 

Let’s dive in :) 

‘Cannes’ You Believe It? AI Takes Over Advertising’s Biggest Week 

At the Cannes Lions Festival of Creativity this year, the AI conversation flipped from “What could we do?” to “Here’s how we’re doing it”. Last year, the conversation was marked by curiosity, and healthy dose of skepticism about AI’s potential; however, in 2025, marketers, agencies, and tech players showed up ready to talk more about how they are implementing Gen AI and moving beyond exploration.  

Meta went beyond just demos and showed the proof in the pudding! The company showcased how their new Gen AI tools, like Advantage+ and Video Generation 2.0, are enabling faster campaign production while keeping creative consistent across channels. The company has consistently demonstrated how it is integrating AI, including Gen AI, to empower marketing and comms teams accelerate time to market while also driving improvements in performance. As a result, the company is capturing a lot of value from agencies with its AI strategy and with how aggressive they have been in hiring talent over the coming weeks, you can expect bigger moves between now and Cannes 2026. 

Meanwhile, Havas became one of the first major global agency networks to formally announce AI agents as "teammates", a clear signal that traditional agencies are embracing Gen AI. The company has already pledged over €400 million toward AI development over four years and CEO Yannick Bolloré says the bet is clear: “AI will help teams find efficiencies, but the bigger play is driving more revenue with the same headcount”. A nicer way of saying “do more work without hiring more people”. Bolloré statement is not new and echo’s what we have heard from the tech sector for much of the last year. 

However, beneath the buzz, how agencies plan to monetize Gen AI surfaced. A Forrester and 4As study presented at Cannes found that 75% of agencies using Gen AI are covering costs out of pocket, up from 41% last year. We have yet to see any major shifts in how agencies are charging clients for work and conversations around efficiency gains don’t seem to be front of mind. 

Amid all the talk of speed and scale, Cannes underscored a timeless truth: storytelling is emotional, and that’s what makes it stick.  

As Apple’s VP of Marketing Communications put it, “The best marketing makes people feel”, while Unilever’s Chief Growth and Marketing Officer reminded us that “desire is emotional, it is not rational”. Gen AI can reduce the friction between idea and execution, but making people feel remains a human superpower, as we have discussed in one of our previous articles. 

Google’s Continues to Experiment with AI Overview in Search 

Google’s AI Overviews have slipped from the top position in 12.4% of U.S. desktop search results, a sharp increase from less than 2% a month ago. This signals that Google is closely watching how people interact with its AI summaries. 

Take this example: if you search “best Gen AI marketing tools”, do you really want a bland AI overview? Probably not. You want a curated, visual, up-to-date list you can scan and act on quickly. Google sees that behavior and where users are skipping AI Overview to click on an organic result, Google is adjusting rankings. In other words, AI overviews at the top of the page for search engine results are not necessarily the new standard.  Depending on the query, you may be served a link first. Your well-optimized, human-centered content can beat Google’s AI if it satisfies the user’s needs more effectively. 

This is a powerful reminder: user behavior shapes search visibility, even in the age of AI. Google is experimenting to find out when AI Overviews help, and when they get in the way of what people are searching. 

Microsoft Stumbles as Nvidia Soars in the AI Chip Race 

Nvidia, the world leader in AI computing, briefly overtook Microsoft last week to become the world’s most valuable company, reaching a market cap of $3.7 trillion, driven by relentless demand for the GPUs (graphics processing units) powering Gen AI models. By comparison, Microsoft stood at $3.65 trillion, despite being a leading AI investor.  

Meanwhile, Microsoft’s Maia AI chip, code-named Braga (both named for Portuguese cities tied to Microsoft’s Azure engineering teams), has been hit by delays, with mass production now pushed to 2026 due to design changes and staffing challenges. This leaves Microsoft trailing Google and Amazon, both of which are already moving their next-generation AI chips into production as all three race to develop in-house chips to reduce their dependence on Nvidia and lower the costs of running Gen AI services. For now, all of them continue to rely on Nvidia’s chips to run Gen AI services (hence Nvidia’s major success). 

Outside of costs, why are all these tech companies working on their own silicon? The connection between performance, models and hardware is so important, that in order to gain a competitive advantage, you need tight connection between models and silicon.  That synergy between model architecture and silicon design is what unlocks the next level of AI capability. By co-designing hardware and software in tandem, optimizing data paths, numerical precision and memory hierarchies for specific model workloads, companies achieve orders-of-magnitude improvements in throughput and energy efficiency compared to off-the-shelf GPUs. This tight integration also future-proofs their platforms: as models evolve, the underlying silicon can be iterated in lockstep, maintaining peak performance, controlling supply chains and cementing a durable competitive edge. But it also creates more risks…

If you can’t keep pace with chip development and design, you risk significant setback. In short, the future of AI won’t be won on model size alone but on the seamless fusion of bespoke chips and tailored algorithms. 

Big Tech Bets on Nuclear to Power AI’s Appetite 

Alongside the chip race, AI’s energy demands are forcing tech giants to rethink power strategies. Generating a single Gen AI image can use as much electricity as fully charging a smartphone and asking a Gen AI model a single question can consume as much power as leaving a light bulb on for 20 minutes... yikes. At scale, the hundreds of millions of AI users worldwide are effectively adding the energy equivalent of millions of new homes to the grid.  

Solar and wind can’t guarantee consistent supply, and fossil fuels clash with climate pledges. Google, Amazon, Meta, and Microsoft are now pursuing nuclear partnerships to keep data centers running 24/7 while meeting emissions targets. 

As AI adoption accelerates, its energy footprint will become a central part of the conversation around ESG commitments. Brands will need to demonstrate how the AI tools powering their daily operations align with their climate goals, as stakeholders increasingly ask, “How green is the AI behind your brand?” in the coming years. 

Courts Begin Defining Gen AI’s Legal Limits 

Last week, we discussed Disney's high-profile lawsuit against Anthropic. Disney is claiming that its copyrighted works were used without permission to train Claude’s AI model. This case reflects a broader wave of litigation testing how copyright law applies to Gen AI’s “training diets”. 

This week, that legal landscape evolved even further. In a separate but related case, a California judge ruled that the outputs from Gen AI models like Claude, can be “exceedingly transformative” and may fall under fair use. This means that AI labs can train their models on copyright materials, such as books, but not infringe on copyright. This is a critical decision because it suggests that the outputs from Gen AI tools are original enough not to infringe or copyright and also qualify for copyright protection themselves. 

This is yet another sign that Gen AI outputs are unique enough, reducing the risks for companies using them for the creation of everything, from content to campaigns.  

The case isn’t over yet. Anthropic will still face trial over allegations that it sourced training data from a large collection of books obtained without authorization, which plaintiffs argue violates copyright and could carry damages of up to $150,000 per work. But the decision suggests we will see more rulings that enable Gen AI companies to operate under fair use while encouraging expansive licensing deals to resolve disputes.  

BBC Launches Gen AI Pilot for Newsrooms 

The BBC launched “Style Assist”, a Gen AI pilot that reformats Local Democracy Reporting Service (LDRS), a network of journalists who cover local councils and public bodies across the UK, to match BBC house style. A trusted LDRS report is submitted, Style Assist drafts a revision, and a senior journalist reviews it before publication. 

The pilot is designed to help the BBC publish more local stories without compromising editorial standards, such as accuracy, impartiality, and clear, consistent house style. Style Assist’s drafts are reviewed by senior journalists, and the BBC plans to disclose when AI has been used, in line with its transparency principles. 

It’s a small but significant signal: trusted newsrooms worldwide are beginning to integrate Gen AI to enhance output while preserving credibility.  

Denmark Moves to Tackle Deepfakes

Last but not least, Denmark announced plans to introduce legislation granting individuals copyright over their likeness and voice to combat unauthorized deepfake use. This approach reflects a growing push across Europe to address the misuse of AI-generated content. 

As synthetic media, such as influencer videos, cloned brand ambassador voices, and virtual product demos, becomes more common in campaigns and customer engagement, brands will need to stay ahead of evolving regulations to protect trust and avoid reputational risks. Denmark may be the first to take this step, but how long before other countries follow? 

Denmark’s move is closely related to the SAG-AFTRA and Narrativ deal in the United States. Under this agreement, actors can license digital replicas of their voices for use in AI-generated ads, with clear terms around consent and compensation. Both examples suggest that if individuals can copyright their voice and likeness, it will undoubtedly reshape how brands engage with cloned content, requiring new approaches to licensing, transparency, and consent in campaigns that leverage Gen AI tools. 

That’s All for Now!

Want to get more out of Gen AI? Schedule a call with us. We’re happy to help kickstart your Gen AI journey! And, if you haven’t subscribed yet, now’s the time. Sign up, share with a colleague—because in the world of AI, sharing is caring. 

Stay tuned for more Gen AI headlines!