Ahead of Times – Issue #7

Veo comes to Google Ads 

Google has rolled out Veo, its most advanced generative video model, directly into Asset Studio within Google Ads. Advertisers can now turn up to three static images into short, high-quality videos of up to 10 seconds, ready to run across YouTube and Demand Gen without leaving the platform, and package them into ready-to-serve ads using customisable templates. 

Worth noting is that Veo works alongside Nano Banana in Asset Studio. While Nano Banana handles image edits such as background swaps, product variations and on-brand refinements, Veo brings those images to life through video. Together, they form a lightweight but powerful product creative system inside Google Ads. 

Why this matters 

Video is no longer a production bottleneck. Brands that have historically relied on static product imagery can now access YouTube and Demand Gen with video assets, without additional tools, or production agencies. Veo makes it easy to generate the level of asset variety Google increasingly rewards, with Demand Gen campaigns performing better when multiple video assets are in play.  

With Veo built directly into Google Ads, the distance between creating, testing and going live shrinks considerably. That said, creative still needs to be reviewed properly. AI-generated video needs to stay on brand and meet the quality bar required, otherwise the efficiency gains come at the cost of brand trust. 

The latest on ChatGPT advertising 

Updates continue to come through, but only in small doses and not directly from OpenAI. Last month the main news was the partnership with Criteo, enabling ads to be bought through its platform. More recently, it has been reported that OpenAI is in talks with The Trade Desk, which would provide another way to access ChatGPT inventory via an established ad platform. 

This could support The Trade Desk’s growth ambitions and help OpenAI demonstrate a clearer path to profitability. Selling ads through platforms such as The Trade Desk and Criteo could help to prove this. While OpenAI is also reportedly building its own ads manager, these early partnerships may be a way to get the ads business off the ground before moving more in house. 

Meanwhile, the ChatGPT ads pilot has expanded to Canada, Australia and New Zealand, and reportedly reached $100 million in annualised revenue in under two months in the US. This includes over 600 advertisers, nearly 80% of which are small and medium sized businesses. 

With $200,000 minimum commitments and CSV based reporting, this is limited for the level of investment. Click-through rates have reportedly been below Google Search and in some cases under 1%, suggesting the need for richer measurement to better understand user quality, intent and potential impact on conversions. 

Why this matters 

ChatGPT is starting to take shape as a new advertising environment, but it is still very early days. The fact that updates are coming through slowly suggests OpenAI is testing demand, partnerships and what the model could look like before fully committing. For advertisers, this points to a potential high intent channel, but one that is still limited when it comes to measurement, transparency and how you actually buy media. It is definitely a space to keep an eye on, rather than rushing in, especially while OpenAI works out how to balance ads, trust, privacy and monetisation. 

OpenAI drops Sora abruptly 

OpenAI has shut down Sora, its AI video generation tool and one of the strongest performers in the market, just six months after the launch of Sora 2. The shutdown has been abrupt. Most Help Centre articles have not been updated, but one quietly changed a few days ago confirms that the Sora web and app experiences will be discontinued on 26 April 2026, and the Sora API on 24 September 2026. 

The timing makes this even more striking. OpenAI was reportedly close to finalising a $1bn deal with Disney on the same day, but the deal did not go ahead. 

Sam Altman explained the decision as a matter of resource allocation, saying OpenAI needed to concentrate its compute and product capacity on the next generation of automated researchers and companies. The coverage across major tech publications points to a few clear factors: 

• Video compute costs were difficult to sustain at scale. Sora was reportedly costing around $1 million per day to run, with no clear path to profitability as usage declined. 

• The platform struggled with safety and moderation issues, including concerns around non‑consensual imagery and copyright. 

• Growing competition from Anthropic and Google appears to have forced a change in priorities. OpenAI’s CEO of applications, told staff that that the company could not afford to be distracted by side projects, with a renewed focus on productivity and business use cases (this means ChatGPT ads.)  

Why it matters: 

Sora’s shutdown shows that even the most hyped AI products cannot survive without a viable business model, no matter how good the technology. For media and content teams who had started building Sora into their workflows or production pipelines, it is a sharp reminder of the risk of building on, and relying on, a single proprietary AI model. When that model disappears overnight, so does everything built around it. 

Having access to a model garden, where your pipelines are not tied to one provider and you can swap based on what performs best, is the way forward. 

It also puts the spotlight on AI governance. Sora never properly resolved the content and legal issues it faced, and that contributed to its downfall. For any organisation using generative AI, having clear frameworks in place is no longer optional.

Source: https://x.com/soraofficialapp/status/2036546752535470382 

OpenClaw and Nvidia step in 

OpenClaw, also known as the ‘Lobster’ because of its logo, has been one of the most disruptive AI tools over the past few months. It became popular quickly as an open-source agentic AI tool. Unlike traditional chatbots, it acts on your behalf, running on your local device, automating tasks and behaving more like an always-on assistant rather than a chat interface. 

However, this comes with significant risks. Allowing a tool this powerful to access local files, emails and systems has raised serious concerns, with multiple vulnerabilities already reported. In response, Nvidia, the most important company in AI computing, has announced the launch of NemoClaw, an agentic AI platform designed to add privacy and security protections to OpenClaw. If effective, this could address OpenClaw’s biggest challenge, security, and determine whether tools like this can move beyond early adopters into wider enterprise use. 

Why it matters: 

Tools like this raise the bar on what AI is allowed to do, and that brings much bigger questions around risk, control and responsibility. When systems can act with little or no supervision, trust becomes just as important as performance. Nvidia’s involvement shows that security is likely to decide how far these tools go beyond early adopters. For businesses, the key takeaway is not about moving fast, but about understanding the boundaries and conditions under which this kind of technology can be used safely. 

Netflix brings AI into filmmaking 

Netflix has acquired InterPositive, the AI filmmaking startup founded by Ben Affleck. InterPositive builds AI tools designed to support filmmakers during production and post production, helping with tasks like visual consistency, edits and effects, without generating content from scratch. Netflix has positioned this as a creatorfirst move, bringing the InterPositive team in house. The focus is on using AI to support creative teams while reducing production costs and speeding up workflows .

Why this matters:  

Netflix is projected to spend around 20 billion dollars on content in 2026, and to help offset this, it is investing in tools that make production more efficient. Bringing InterPositive in house shows how Netflix plans to use AI to reduce costs and speed up post production, while keeping creative control and judgement with people. The goal is not to replace creatives, but to make the process more efficient, with AI supporting the work and humans making the final decisions, an approach brands should be following too. 

ChatGPT shopping changes direction 

OpenAI has reportedly scrapped its Instant Checkout feature in ChatGPT after finding that users were mainly using the tool for research and discovery, rather than completing purchases within the chat. Instant Checkout was designed as an end-to-end shopping experience inside ChatGPT, allowing users to buy without leaving the conversation. 

Instead, OpenAI is now shifting purchases to third-party apps connected to ChatGPT, where brands and retailers manage checkout in their own platforms. ChatGPT will focus on helping users discover and compare products, rather than acting as the point of sale. 

Why this matters:  

This reinforces how people are actually using ChatGPT today, mainly for research and discovery rather than completing transactions. Longer term, this does not rule out a full end to end commerce experience inside ChatGPT. It reflects the fact that the technology, infrastructure and frameworks around security, trust, accountability and payments still need to mature. As those challenges are addressed, seamless purchasing within AI tools may become more viable. For now, OpenAI is focusing on what works, discovery and intent, while leaving transactions in trusted third party environments.