Week In Review: It’s been a big week for OpenAI as it launches a consultancy arm and ChatGPT’s personal assistant agent, Netflix uses AI effects for the first time to speed up TV production, and Cognition AI acquires code editing startup Windsurf following Google’s licensing deal.
Brands and Agencies
Popeyes Releases AI-Generated Diss Track
Louisiana-founded chicken restaurant Popeyes released an AI-generated diss track against competitor giant McDonalds. The ad was created with help from AI filmmaker PJ Accenturo, and centres around McDonalds announcing the re-launch of its popular Snack Wraps a day after Popeyes announced its own Chicken Wraps. PJ Accenturo reportedly used AI music production tool Suno and Google’s video-generation platform Veo 3 to produce the ad. Accetturo described on X: “We started with image-to-video, but it was too slow and we had less than 3 days when I scrapped all of our work and said we would use only Veo 3. We wanted something that went hard but didn’t take itself seriously. Once we switched to Veo 3, creating a story about people having fun around town with chicken wraps became easy. It all came together when we added the clown bits and the crazy animals at the party.”
Matt McNeany to Lead WPP Open
WPP has brought in former Omnicom-exec Matt McNeany to lead its new AI operating system WPP Open. Details of the appointment haven’t been shared, however McNeany updated his LinkedIn profile, confirming the move. He served as Chief Transformation Officer at Omnicom Precision Marketing Group (OPMG). WPP Open was launched in June this year and combines four proprietary AI models – ‘performance brain’, ‘brand brain’, ‘channel brain’, and ‘audience brain’.
Vodafone Announces Success of AI-Powered Brand Suitability
Telecommunications company Vodafone announced that implementing AI into its brand strategy opened its ads up to 10 percent more news publisher inventory. The company utilised brand suitability controls developed by media measurement company DoubleVerify. Using DoubleVerify’s technology, Vodafone also reported a 41 percent decrease in ad blocks from 2024 to 2025 and a 48 percent decrease in keyword violations year-over-year. DoubleVerify’s intelligence system uses machine learning algorithms to modify ads to fit trends and adhere to brand guidelines. The technology analyses video, audio, image, and text content. Vodafone decreased brand suitability violations by 46 percent between 2024 and 2025 and its engagement from news platforms rose 11 percent above baseline toward the end of 2024.
IAB Report Reveals 90% of Advertisers Will Use AI for Video Ads
The Interactive Advertising Bureau’s 2025 Digital Video Ad Spend & Strategy Report shared that 90 percent of advertisers will use AI tools to create video ads by 2026. The report also included that 86 percent of ad buyers currently use or plan to use AI for video ad creation. It is predicted that 40 percent of all ads will have used AI by 2026 and that CTV inventory will be 47 percent biddable, likely as a result of AI deployment. More advertisers are turning to AI for ad creative, as the tools offer more cost efficient production avenues.
ChatGPT Will Let Users Purchase Products In-App
OpenAI, the parent company of ChatGPT, is working on letting users of its popular Gen AI tool shop within the chatbot interface, according to The Financial Times. The AI giant has been looking for ways to expand its business model, which primarily relies on income from user subscriptions. In April, the company announced it would be partnering with e-commerce platform Shopify, which fuels TikTok Shop – the Chinese social platform’s in-app retail space. The offering will fundamentally change the way people shop, and the way brands engage with customers.
Media
Netflix Uses AI Effects For First Time
Streaming giant Netflix revealed it has used AI-generated visual effects in one of its TV shows for the first time. One of the company’s CEOs Ted Sarandos said AI was used to create a scene of a building falling down in “The Eternaut”, one of its science fiction TV series. Sarandos said that AI-use has helping the streaming company to create productions faster and at lower costs, with the building scene being created ten times faster than with traditional methods. He said: “That sequence actually is the very first [generative] AI final footage to appear on screen in a Netflix original series or film. So the creators were thrilled with the result.”
UK AI and Creative Industries to Develop Copyright Solutions
The UK’s Technology and Culture Secretaries began working on an AI innovation plan that maintains protection for creatives from AI companies. Meetings will be attended by representatives from the AI and creative industries, including from Meta, Amazon, OpenAI, Sony Music Entertainment, and The Guardian. The meetings will begin with discussions surrounding opportunities with AI and copyright disputes. These new working groups are part of the UK’s Plan for Change, a set of milestones introduced by Prime Minister Keir Starmer in December 2024. The UK has seen intense clashing between the industries, with disagreements over copyright regulations for AI developers.
Google Adds AI Summaries to News Feature
Google Discover, the news feed on Google’s app, will now include AI Overviews. Similar to what appears with searches on its engine, users will get a summary at the top of the page based on reports from various sources. Previously, the platform included full headlines and publication names. It will now only showcase logos of the companies it refers to as a source. Google provided an example of what this will look like in practice, showing a story about US President Donald Trump and Ukraine offering links to similar articles about Trump. This feature will be launched for US users. The AI Overviews are not currently generated for all news stories, and are mainly shown for lifestyle-related stories.
Google Partners with Publishers for Featured Notebooks
Google is releasing specified notebooks within NotebookLM, its AI note-taking tool. These “featured” notebooks will be already filled with information on selected topics. This includes a notebook with research “The Complete Works of William Shakespeare” and “An overview of long-term trends in human well-being” from Oxford University research. Google is partnering with publishers The Economist and The Atlantic to create additional featured notebooks. NotebookLM generates summaries on the information for each notebook, creating notes for the user to review. Users can then ask the app’s chatbot questions, and listen to “Audio Overviews”, AI-generated “podcasts” about the topics.
Tech
OpenAI Launches Personal Assistant Agent
OpenAI launched its own personal assistant agent within ChatGPT, that can go shopping and make restaurant reservations on a user’s behalf. “ChatGPT agent,” can conduct mass research and take control of web browsers, in addition to desktop files like spreadsheets. In a statement, OpenAI said the assistant can handle requests like “look at my calendar and brief me on upcoming client meetings based on recent news” and “analyse three competitors and create a slide deck.” The new offering brings together some of ChatGPT’s existing functionalities, such as Operator’s ability to freely navigate websites, thorough analysis from deep research, and the platform’s natural language functionality.
AI Companies Unprepared for AGI, Says Safety Group
The Future of Life Institute (FLI), a prominent AI safety group, has warned that AI companies are “fundamentally unprepared” for the repercussions of creating human-like intelligence or artificial general intelligence (AGI). The FLI’s index reviewed all major LLM providers, including Google DeepMind, OpenAI, Anthropic, Meta, and DeepSeek, Zhipu AI, and Anthropic. The group said all companies scored lower than a D for “existential safety planning” in ensuring AGI would remain safe. However, Anthropic received an average rating of C+ – the highest overall safety score.
OpenAI Launches Consulting Business
OpenAI announced the launch of a consulting arm, starting at $10m for the service. According to The Information, the AI giant will offer bespoke consulting services, with major clients reportedly including the US Department of Defense and a Southeast Asian-based app called Grab. OpenAI will create company-tailored versions of its GPT models and place its engineers within organisations to help integrate the technology. The offering is being led partly by researcher and computer scientist Aleksander Mądry, who is focused on building LLMs to meet customer needs. In preparation, the giant has hired engineers from major consultancy group Palantir.
Scale AI Cuts 14% of Workforce
Data labelling company Scale AI has cut 14 percent of its workforce amid significant investment from Meta, and the hiring of the firm’s CEO. Meta invested $14bn into the AI company and hired CEO and founder Alexandr Wang as Meta’s new chief AI officer. Off the back of this, Scale AI has placed Jason Droege as an interim CEO and cut 200 employees. In a statement to CNBC, the company said “This structure will allow us to better serve the customers we have today and win back customers that have slowed down work with us.”
Apple Reportedly Considers Buying Mistral
Apple is reportedly toying with the idea of buying AI startup Mistral. According to reports from Bloomberg, Apple, widely seen as lagging in the AI race, is considering acquiring the French startup that has raised a total of $1.1bn over seven funding rounds. Mistral is Europe’s biggest AI company, and is valued at $5.5bn. The news is part of wider speculation on Apple acquiring an AI company to bolster its capabilities, with recent rumours about the giant purchasing US-based startup Perplexity AI.
Meta Acquires Voice Generation Tool PlayAI
Meta settled a deal with PlayAI, an AI voice generation startup. All of PlayAI’s staff will be joining Meta as part of the deal. Meta shared hopes of using PlayAI’s technology to boost the abilities of its AI assistant and the chatbots on social media platforms like Instagram and Facebook. The company reportedly said that PlayAI’s natural voice creation abilities would be useful in advancing Meta AI characters, Wearables, and audio content creation. This deal comes as Meta makes significant investments in AI. The tech company also recruited staff from other tech giants like OpenAI and finalised a $14bn deal with startup Scale AI to improve model training.
Alibaba’s Moonshot Releases New Large Language Model
Moonshot, the Alibaba-backed startup, came out with a low-cost large language model (LLM) that challenges OpenAI’s ChatGPT. The model, Kimi K2, is particularly skilled in writing code. Moonshot shared on social media that the model had better performance than OpenAI’s coding model and that it surpassed Anthropic’s Claude Opus 4 model on two benchmarks. The LLM is open-source, meaning its code is publicly accessible. Releasing lower-cost, open-source models is not common in the western tech space. DeepSeek’s notable models released earlier this year were also low-cost and open-source. Kimi K2 is available for free on the Kimi app and browser.
Meta to Invest Billions into US Data Centres
Meta chief executive officer Mark Zuckerberg announced the company plans to put hundreds of billions of dollars toward building data centres in the United States. The first data centre is called Prometheus and is planned to appear online in 2026. The site will be in New Albany, Ohio. Zuckerberg added that one of the data centres will have a surface area close to Manhattan’s, about 60 square kilometres. This “cluster”, called Hyperion, could eventually generate up to five gigawatts, Zuckerberg said. Following the announcement, Meta’s shares grew 1 percent.
Copilot Update Gives AI a Full View
Microsoft is releasing an update to its AI assistant, Copilot, on Windows. Copilot Vision allows the tool to see everything the user is doing on their computer screen. The AI could previously only view two apps at once. Users select which desktop they want Copilot Vision to view. The AI tool then can generate insights, analyse content, and answer questions, according to Microsoft. It can also give advice on improving resumes or creative projects, or assist with a game the user is playing. The feature operates like screen sharing on a video call, and users can stop the screen share by pressing “Stop” or “X”.
AI Startup Thinking Machines Lab Raises $2bn
Thinking Machines Lab, the AI startup created by former OpenAI chief technology officer Mira Murati, raised $2bn in fresh capital. The startup plans to release its first product in the next few months. Investors included Nvidia, Accel, ServiceNow, Cisco, and a16z. Murati left OpenAI in September and introduced Thinking Machines Lab in February. She served as the interim chief executive officer at OpenAI while chief executive Sam Altman was ousted by the company. Murati shared the startup’s early developments will include an open-source tool which can be used by others in the AI industry.
Cognition AI Acquires Windsurf
Startup Cognition AI acquired Windsurf, the startup focused on coding assistance for a price not yet disclosed. This acquisition follows Google’s $2.4bn deal with Windsurf that involved its gain of Windsurf chief executive officer Varun Mohan and co-founder Douglas Chen and a rumoured potential acquisition from OpenAI. Cognition will acquire Windsurf’s product and intellectual property along with the remainder of its staff. Cognition plans to incorporate Windsurf’s technology into its own products. Windsurf will continue to develop its AI integrated development environment (IDE), a resource for AI developers, and Cognition will continue work on Devin, an AI-powered coding assistant and software engineer.
Meta Fixes Chatbot Bug That Raised Security Concerns
Meta fixed a bug in its AI chatbot which allowed others to view queries and responses from user conversations. The bug was identified by Sandeep Hodkaisa, founder of security firm AppSecure. After reporting the bug in December 2024, Meta paid him $10,000. Hodkaisa analysed network traffic in his browser and found he could receive an AI response that had been given to a different user previously. This also meant that Meta’s servers did not have adequate guardrails in place to ensure the responses were being distributed to the right users. Meta fixed the bug on January 24, 2025, and stated it found no evidence that the bug had been used to steal or leak any data.
Google to Roll Out New Search Features
Google announced new AI features to come to Search. Gemini 2.5 Pro, Google’s advanced reasoning model, will be integrated into “AI Mode.” This lets users get answers to more complex queries and get math and coding assistance. AI Mode will also be integrated with Deep Search, Gemini 2.5’s research tool. With Deep Search, users can generate a fully-cited report on a topic that interests them. Google added that Search will also be getting new agentic capabilities – a feature that calls businesses on behalf of the user. In action, this may look like the user searching “pet groomers near me” and selecting the “Have AI check pricing” option. The tool will then call businesses, collecting information to provide the user various appointment times and service options. Google Search’s AI Mode features will be available for Google AI Pro and AI Ultra subscribers in the US, and the agentic features will be available for all US users.
AWS Releases Agentic AI Tools
Amazon Web Services (AWS) announced the release of new agentic AI tools at the AWS Summit in New York. This includes the launch of Amazon Bedrock AgentCore, a platform to build and deploy AI agents. AgentCore Identity, one of the platform’s services, ensures security protection by offering agent authentication. Other services include a code interpreter tool, a tool for integrating AI agents into browsers, and tracking an agent’s actions. In addition to AgentCore, Swami Sivasubramanian, AWS vice president for agentic AI, revealed a set of agentic AI tools for AWS Marketplace. AWS Marketplace, a store for AWS software and services, will have AI agents and tools to help customers “discover, buy, deploy, and manage AI agents and tools from leading providers”. AWS added it will be investing an additional $100mn into its AWS Generative AI Innovation Center. Having already invested $100mn in the centre after its opening in 2023, the company hopes to further “accelerate customers’ development of autonomous, agentic AI systems”.
Nvidia to Continue Shipping H20 Chips
US chip giant Nvidia has revealed it will resume selling its H20 chips to China, following a ban imposed by President Donald Trump’s administration in April. The move reverses a ban on AI chips being sent to China amid concerns the tech will be used to advance potentially dangerous AI models. After previous export restrictions outlined by the Biden administration in 2023, Nvidia created the H20 chip specifically for the Chinese market. However, was told earlier this year by the current US administration that even this chip would no longer meet export rules. However, in a blog post, Nvidia said the US government has given the company license to restart sales of its Chinese-specific H20 chips in the region. Nvidia’s CEO Jensen Huang has met with US policymakers and Chinese officials for the chips to be sold in China once more.
Mistral Reveals Speech Recognition Model
French startup Mistral released Voxtral, its first set of audio models for enterprises. Voxtral offers audio tools at “less than half the price” of competitors. Voxtral transcribes audio up to 30 minutes long, and understands audio up to 40 minutes. It relies on the large language model (LLM) Mistral Small 3.1 for its reasoning capabilities. This LLM allows the model to answer questions, generate summaries, and use voice requests to carry out actions. The model automatically detects and uses multiple languages, including English, Spanish, French, Hindi, Italian, and Dutch.
Number of the Week
$25 billion. That’s how much Google said it would invest in data centres and AI infrastructure in the US over the next two years. The development will occur within the PJM Interconnection area, the largest power grid operation site in the US. The grid covers 13 states in the mid-Atlantic, Midwest, and southern regions. Google will also invest $3bn into refurbishing two hydropower plants in Pennsylvania to keep up with data centre demand in the state. The hydropower plant investment is part of a deal Google signed with investment company Brookfield Asset Management, which involves the purchase of 3,000 megawatts of hydroelectric power in the US. US President Donald Trump met with White House Cabinet officials and tech executives to discuss AI development in Pennsylvania. The president praised companies for an amassed $90bn investment in AI infrastructure in Pennsylvania alone.
