OpenAI's Code Red, Voice AI's $70M Bet, and the Race to Optimize for ChatGPT

Welcome to today's edition of AI Business Weekly. From OpenAI's internal "code red" as ChatGPT faces its fiercest competition yet to a Paris voice AI startup securing $70 million just three months after founding, this week reveals an industry simultaneously defending established positions and racing toward new frontiers. A Canadian city deploys AI across 3,200 municipal employees while researchers expand AI training tools for suicide prevention conversations, and Berlin's Peec AI triples its valuation in four months as brands scramble to optimize for ChatGPT instead of Google. These developments underscore a critical evolution: artificial intelligence is no longer just a technology race—it's becoming embedded infrastructure across government operations, clinical training, marketing strategy, and the very interfaces through which humans and machines communicate. The question is no longer whether organizations will adopt AI, but how quickly they can adapt as the competitive landscape shifts beneath their feet. Let's dive in.

OpenAI CEO Sam Altman issued an urgent internal memo declaring a "code red" to refocus company resources on improving ChatGPT's core experience, exactly three years after its revolutionary launch. The directive comes as Google's Gemini 3 has beaten ChatGPT on industry benchmarks and Anthropic's Claude Opus 4.5 gains enterprise traction, forcing OpenAI to delay advertising plans and other projects. Despite a $500 billion valuation and 800 million weekly users, OpenAI isn't profitable and has committed over $1 trillion to cloud providers, making ChatGPT's market leadership critical to achieving projected revenue growth from $13 billion in 2025 to $200 billion by 2030. Read more

Paris-based Gradium emerged from stealth with a $70 million seed round from investors including former Google CEO Eric Schmidt and French billionaire Xavier Niel, just three months after its September 2025 founding. Founded by former Google DeepMind researcher Neil Zeghidour, the startup has developed ultra-low latency voice AI using audio language models that work directly with sound rather than converting speech to text and back. The company launched with multilingual support in five languages and reportedly started generating revenue within weeks of founding, positioning itself to compete against OpenAI's voice capabilities and specialized players like ElevenLabs. Read more

Now through the end of the year is prime streaming time on Roku, with viewers spending 3.5 hours each day streaming content and shopping online. Roku Ads Manager simplifies campaign setup, lets you segment audiences, and provides real-time reporting. And, you can test creative variants and run shoppable ads to drive purchases directly on-screen.

Bonus: we’re gifting you $5K in ad credits when you spend your first $5K on Roku Ads Manager. Just sign up and use code GET5K. Terms apply.

Use code GET5K now

The City of Ottawa revealed that 3,200 municipal employees are now using Microsoft Copilot, having submitted over 114,000 AI prompts for tasks like finding information and drafting communications. Officials announced new AI projects for resume screening in job applications and automatically identifying streetlights and signs in footage from the city's mapping vehicle, while emphasizing the technology augments rather than replaces workers. City Manager Wendy Stephanson confirmed a comprehensive AI governance framework covering privacy and responsible use will launch in 2026, following similar initiatives by Ontario's provincial government where over 15,000 civil servants use Copilot weekly. Read more

New funding from USAA, the Humana Foundation, and Reach Resilience will support development of two AI training programs focused on firearm safety discussions and crisis response planning for mental health clinicians treating veterans and military personnel. The tools, expanding UT Health San Antonio and Rush University's STRONG STAR Training Initiative, allow clinicians to practice sensitive conversations in realistic AI-simulated scenarios anytime, building confidence before engaging with real patients. Since firearms are used in roughly 72 percent of veteran suicides, secure storage conversations are vital but often difficult for clinicians, with early feedback on existing AI tools being overwhelmingly positive. Read more

Berlin-based Peec AI closed a $21 million Series A just four months after its seed round, with its Generative Engine Optimization platform helping brands maintain visibility as consumers increasingly ask ChatGPT instead of Google for recommendations. The company's valuation has tripled to above $100 million, driven by growth to over $4 million in annual recurring revenue within ten months, serving 1,300 companies including Chanel and ElevenLabs. The platform tracks prompts rather than keywords, showing brands which sources influence AI-generated answers and providing actionable recommendations to improve visibility and sentiment in ChatGPT, Perplexity, and Google Gemini results. Read more

Developments expose how quickly competitive advantage evaporates in AI. OpenAI's "code red" three years after launching ChatGPT shows that first-mover advantage means nothing when Google and Anthropic close capability gaps in months. Meanwhile, Peec AI's $21 million raise reveals that AI's disruption immediately spawns secondary markets—as ChatGPT replaces Google for discovery, Generative Engine Optimization becomes its own venture-backed category. The contrast between Ottawa deploying AI across 3,200 employees and researchers carefully expanding suicide prevention training illustrates AI's bifurcation: mature enough for municipal productivity gains, sensitive enough to require extreme caution in clinical applications. Gradium's $70 million seed round signals voice will dominate AI interfaces, but the funding size also shows how expensive competitive infrastructure has become. The pattern isn't any single story—it's their juxtaposition. AI is simultaneously mature enough for production deployment and nascent enough that fundamental interfaces are still being invented. Winners won't just build the best models; they'll correctly identify which layer of the stack offers defensible advantage as technology stabilizes into infrastructure.