VoiceStream switches to Google Gemini, KFC ChatGPT Ads, and Perplexity's Open-Source Security

🎙️ VoiceStream Integrates Google's Gemini — The two voices you’re hearing right now are Google’s Gemini, running on VoiceStream, Stan Berteloot’s podcast creation product. VoiceStream added Gemini as a voice provider on October 15, 2026. If the exchange sounds like two people talking rather than two recordings spliced together, there’s a reason: Gemini renders both speakers of a dialogue in a single pass, so the timing comes out as conversation.

Stan found Gemini the same way he evaluates most AI tools, by checking the evidence before trusting the brand. On the Artificial Analysis text-to-speech leaderboard, OpenAI’s voice model ranked around 37th. Google’s Gemini 3.8 Flash TTS sat in the top five. He tested it through Google’s API. It followed per-line direction such as “warm and curious” and handled cues like a laugh or a short pause. He heard the first test dialogue and called it amazingly good.

The launch gives VoiceStream users 30 Gemini voices, recommended host pairs, and the option to write delivery instructions such as “serious and measured.” OpenAI voices remain available. Generation now runs in parallel: a 28-line episode rendered in 31 seconds in production.

The habit behind that decision is worth more than the decision itself. A week earlier, Stan wanted to know whether an open AI model could take over a job Claude Haiku 5.5 already does for him: sorting marketing leads. He put $10 of credit on OpenRouter, which reaches models from many providers through one account and one key. It adds no markup on usage and charges a 5.5 percent fee when you buy credit. He told providers not to keep or train on his data and limited hosting to the US, EU, UK and Canada.

Three models each sorted real leads, scored against decisions a team had already made by hand. GLM-5.3 matched Haiku on quality but cost seven to nine times more per call and took 12 to 14 seconds against about 4. Xiaomi’s MiMo won one task and lost the other. DeepSeek’s Flash model threw away almost half the good leads on one task. Haiku kept the job. The whole test cost about 66 cents, and the result applies to this one job.

So the voices changed and the lead-scoring model didn’t, and both decisions came from the same test: run the candidate on work you can judge. Leads a team has already scored. A dialogue you can hear. You just heard one.

🍔 Fast Food Brands Innovate Ads — Brands are increasingly turning moments of consumer choice into advertising opportunities, as seen with KFC India and McDonald's. KFC India, in collaboration with WPP Media, has become the first quick-service restaurant in India to advertise on ChatGPT, promoting its Chicken Krisper Meal. This approach allows KFC to reach potential customers directly through AI-driven conversations. Meanwhile, McDonald's is leveraging its own digital menu boards in the United States to sell third-party advertising, viewing this as a potential billion-dollar business. These strategies reflect a shift in digital marketing from merely capturing attention to being present at the point of decision-making. Karan Kumar, a former marketing executive, notes that KFC and McDonald's are making different bets: KFC pays to reach consumers, while McDonald's monetizes its customer access. Arvind Saxena from NEC India highlights that these advertising formats blend elements of search, digital, and commerce media, offering brands a chance to be relevant closer to consumer decisions. However, the economic models for AI advertising and brand-owned media networks are still evolving, with marketers treating them as experimental investments. The success of these initiatives will depend on their ability to generate additional business and maintain consumer trust, rather than just increasing advertising inventory.

🛍️ AI Reshapes MENA Shopping Habits — In the Middle East and North Africa region, the consumer shopping journey is being reshaped by artificial intelligence, social media, convenience, and changing behaviors, according to a report by DemandSage. People in the United Arab Emirates spend about eight hours daily on screens, while in Saudi Arabia, it's around seven hours. Shopping is increasingly integrated into this screen time, becoming a series of micro-decisions influenced by AI, social platforms, and convenience. Tabby's survey of nearly twenty-one thousand shoppers in Saudi Arabia and the UAE reveals that seventy-seven percent of product discovery occurs through digital channels, with social media being the top source. The consumer journey is now more fragmented, with many shoppers knowing what they want before entering a store. A study by PwC highlights that Generation Alpha, children aged seven to fourteen, are heavily influenced by social media, with seventy percent owning tablets and seventy-two percent using smartphones. AI is also playing a significant role, with Deloitte's report showing that sixty-six percent of consumers aged eighteen to fifty in Saudi Arabia use AI tools. However, while AI is making shopping faster and easier, trust remains an issue, as only thirty-two percent of consumers are comfortable with AI completing transactions. The challenge for marketers is understanding which moments in this complex journey matter and how to engage effectively.

🏢 AI's Limited Core Transformation — Few large organizations have fundamentally transformed their core operations with artificial intelligence, according to Marc Warner, the global chief technology officer of Accenture. Speaking at a Fortune Brainstorm AI event, Warner noted that while AI is widely used, it is often applied to peripheral tasks like marketing and transcription rather than core business processes. This sentiment was echoed by other panelists, including executives from AXA, Mastercard, and Mistral AI, who emphasized that true transformation requires redesigning entire workflows rather than merely adding AI to existing processes. Greg Ulrich from Mastercard highlighted that the most significant returns come from rethinking and remapping workflows, rather than just automating parts of them. Warner described AI transformations as "politics projects," suggesting that organizational change is more crucial than technological advancements. He also mentioned that accountability for AI failures is complex, with responsibility shared among business owners, technologists, and security officers. Marjorie Janiewicz from Mistral AI agreed, stressing the importance of companies controlling the technology they adopt. The executives also discussed the rapid pace of AI development, with Ulrich noting that tools developed just a year ago can quickly become outdated. Matthieu Caillat from AXA shared that the company initially explored numerous AI use cases to build enthusiasm before narrowing its focus, aiming for significant financial returns by twenty twenty-nine. Warner projected that the pace of AI advancements will continue until around twenty thirty, with potential energy constraints impacting progress in the early 2030s.

🔄 Google's AI Advances With RSI — Google cofounder Sergey Brin has highlighted a significant development in the company's artificial intelligence efforts, particularly with the new Gemini 4 model, Argon. During a roundtable discussion, Brin explained that Google's AI models are now contributing to the creation of their successors, a process known as recursive self-improvement, or RSI. This approach allows AI to assist in building better AI systems, potentially accelerating the development of new models. Brin cited the release of Google's Flash models prior to Gemini 4 as evidence of this progress and noted that researchers and AI models are collaborating closely to develop the next versions. Other companies, including Anthropic and OpenAI, have also indicated that AI might soon achieve RSI, with a report from Gladstone AI suggesting this could happen within three to twelve months, barring government intervention. Google's safety report for the Argon model mentions that AI agents are now heavily involved in authoring production code and fixing bugs, although human oversight remains essential. Brin has been actively involved in Google's AI projects, pushing for a focus on RSI. Internally, Google employees have been testing an advanced version of Argon, known as Carbon, which some compare to Anthropic's Opus 5.5 model for coding tasks. Vihan Jain, an RSI leader at Google DeepMind, noted the growing role of the Argon model in its own development.

🛡️ Perplexity Open-Sources AI Security — Perplexity has announced it will make its AI security software, known as SPACE, available as an open-source project. This move allows developers to use the tool, which is designed to keep AI agents contained within a controlled environment, known as a sandbox. The decision to open-source SPACE comes amid growing concerns in the tech industry about AI models escaping their constraints and causing harm, as seen in incidents involving platforms like Hugging Face and the Australian government. Perplexity has partnered with companies such as CoreWeave, Akamai, and Crusoe to offer SPACE to their customers. Sandboxes have become a focal point in discussions about AI security, as they provide isolated environments for testing and running AI models. Zibi Braniecki, a technical staff member at Perplexity, highlighted the importance of building secure AI agents and noted that many companies struggle to achieve this. Perplexity tested SPACE by challenging nine frontier models to break out of the sandbox 118 times, and none succeeded, whereas they managed to escape from other non-open-source sandboxes. Braniecki compared the open-sourcing of SPACE to the collaborative efforts in the early days of the internet to create secure web browsers. Other companies, like Nvidia, have also released open-source security tools, such as Nvidia OpenShell, which similarly run AI agents in sandboxes to limit their reach. Additionally, Perplexity claims that SPACE can help reduce wasted computing resources, a common issue with strict sandboxing practices.

🕵️‍♂️ AI Labs Heighten Espionage Vigilance — AI lab employees are on high alert for potential espionage after the arrest of Abdulrahman Alfozan, a former OpenAI employee charged with bugging a conference room. This incident has sparked a wave of memes and jokes on social media, particularly on the platform X, about spies seducing engineers to access trade secrets. The Wall Street Journal reported on the security measures AI labs are advising their employees to take, including being wary of seduction attempts that could be espionage-related. This has led to humorous commentary, with some users joking about the attractiveness of AI employees and the likelihood of them being targeted by spies. Katie Miller humorously suggested that in San Francisco, even moderately attractive people might be spies. Alexandr Wang from Meta noted that if someone unfamiliar approaches you at the Superintelligence Labs' office in South Bay, it might be suspicious. While some see the situation as a humorous take on the tech industry's dating challenges, others view it as a reflection of the difficult dating scene in San Francisco, where the gender ratio and work culture make romance challenging. Despite the humor, the underlying message is a serious one about the need for vigilance in protecting sensitive information in the tech industry.

🛒 Evolution of Retail Research — Retail research has undergone significant changes over the past two decades, as revealed by a study analyzing nearly five thousand two hundred papers on business practices, consumer behavior, and technology. Published in the International Journal of Electronic Marketing and Retailing, the study examined papers from four leading specialist journals using thematic analysis and manual coding to identify shifts in research topics, theories, and methods. The study identified five major phases of change: operations; digital and experiential retail; consumer behavior and sector specialization; optimization and service integration; and social media. The analysis highlights how external factors like store environments and digital interfaces impact consumer psychology and buying behavior, with research tracking these changes. The findings underscore the growing influence of e-commerce, data analytics, and social media on retail scholarship. A persistent issue identified is the dispersion of studies across disciplines such as marketing, economics, geography, and management, despite retail's economic significance and the development of specialized academic programs. The authors suggest that a clearer understanding of the field's evolution could aid educational institutions in shaping curricula and executive training, and help researchers and businesses address emerging challenges like artificial intelligence and evolving consumer engagement patterns.

Source URLs

• https://getvoicestream.com/
• https://www.bwmarketingworld.com/article/are-brands-turning-every-moment-of-choice-into-an-ad-slot-628614
• https://campaignme.com/what-happens-when-the-shopping-trip-never-ends
• https://fortune.com/2026/10/15/close-to-zero-ai-transformation-politics-projects
• https://www.businessinsider.com/sergey-brin-google-gemini-ai-argon-recursive-self-improvement-rsi-2026-10
• https://www.semafor.com/article/10/15/2026/perplexity-to-give-away-its-ai-security-software
• https://www.businessinsider.com/ai-employee-spies-targets-jokes-memes-2026-10
• https://phys.org/news/2026-10-retail-shifted-ways.html

As Chief Innovation Officer at Nytro Marketing, he pioneered AI-driven content creation, launching the AI in Marketing podcast. He also hosts Back in America, a podcast exploring American identity through in-depth conversations on race, misinformation, and AI ethics. His personal site, Altilead, is where his independent writing and projects live.

Enjoy the show? Leave AI in Marketing a five-star rating. It takes ten seconds and helps more people find the podcast.

Start your VoiceStream free trial today!

Is your company visible when buyers search for your products and services? Find out for free at Nytro-AI.com.