Catch up on select AI news and developments from the past two weeks (in no particular order):
MIT study finds 95% of AI projects deliver zero returns. A new MIT study analyzing 300 public AI initiatives found that 95% of organizations studied gained no return on AI investments, despite $30–40 billion in enterprise spending. Companies buying AI tools fared better than those building internal pilots, but most initiatives failed to produce disruption. The findings shook investors already wary of Big Tech's record AI spending. Analysts warn the results could undermine Wall Street's patience with AI spending, sparking concerns of an AI investment bubble similar to 2000's dot-com collapse. Importance for marketers: The study underscores the need to focus on AI projects that deliver measurable ROI, not just experimentation, as investor skepticism could reshape available tools and partnerships.
AI thought leader warns of catastrophe as global race accelerates. Carnegie Mellon's Anand Rao warns that unchecked global AI investment could create societal catastrophe. With the US, EU, India, and China all pledging billions for AI infrastructure, Rao argues Big Tech may win today's arms race but lose the war by displacing workers, worsening emissions, and undermining trust. He cites acute skills shortages, gender disparities in AI training, and surging emissions from AI data centers as looming crises. Rao suggests an aviation-style safety model for AI regulation, balancing innovation with trust and accountability. Importance for marketers: This critique highlights risks of overhyped AI adoption without safeguards, pressing marketers to adopt responsible AI practices to avoid reputational fallout.
Meta restructures AI unit for the fourth time in six months. Meta is undergoing its fourth AI restructuring in six months, splitting its Superintelligence Labs into four groups: TBD Lab, products, infrastructure, and FAIR research. CEO Mark Zuckerberg is intensifying efforts toward artificial general intelligence, raising annual capex forecasts up to $72 billion to fund massive data centers. The reshuffle follows staff exits and weak reception of its Llama 4 model. Financing includes a $29 billion bond deal for a Louisiana data center. Costs for talent and infrastructure are projected to outpace 2025 growth. Importance for marketers: Ongoing instability at Meta's AI group could impact future AI-driven ad tools and platform capabilities that marketers depend on.
Google debuts Pixel 10 phones with AI-powered coaches. Google unveiled its Pixel 10 lineup, adding AI features including a photography coach and a health coach integrated into Fitbit Premium. All devices feature the new Tensor processor and Gemini Nano AI. The phones also embed industry-standard content credentials into photos to verify authenticity, a first for smartphones. Prices range from $799 to $1,799. The launch aims to outpace Apple's September iPhone event. Importance for marketers: AI integration in consumer devices shapes expectations; authenticity credentials could influence brand trust in digital content and campaigns.
Google expands AI Mode in Search globally with agentic features. Google is expanding AI Mode to 180 countries, now offering chatbot-style queries and agentic features like restaurant bookings through partners including OpenTable and Resy. Premium Google AI Ultra subscribers can use AI Mode for reservations by specifying time, place, and cuisine. In the US, AI Mode is testing personalization based on past searches. Users can also share AI-generated search results via links. Importance for marketers: The integration of booking and personalization into search highlights a new era of AI-driven consumer engagement, altering discovery and purchase pathways.
Study shows older Americans increasingly adopt AI tools. A University of Michigan survey of 3,000 adults over 50 found that 55% had used AI tools, with voice assistants like Alexa dominating over chatbots like ChatGPT. Nearly one-third use AI-powered home security devices, with 96% reporting feeling safer. While 54% trust AI, confidence in spotting inaccuracies is low, with education level influencing adoption. Nine in ten want AI labeling in search results. The findings reveal growing interest but concerns about overtrust and privacy. Importance for marketers: Older demographics are adopting AI, creating opportunities for tailored products and campaigns while emphasizing trust and transparency.
Nvidia releases open dataset for 25 European languages. Nvidia launched Granary, a 1-million-hour multilingual dataset covering 25 European languages, plus new Canary and Parakeet models for speech translation. The dataset, built with Carnegie Mellon and Fondazione Bruno Kessler, enables high-quality AI for underrepresented languages like Maltese and Estonian. Canary offers high accuracy, while Parakeet emphasizes speed. Both are optimized to run efficiently on devices. Importance for marketers: Expanded multilingual AI capabilities lower barriers to reaching European audiences, opening new opportunities for global campaigns with accurate translation and voice tools.
Meta launches AI-powered voice translations for creators. Meta rolled out global AI-powered translations for Facebook and Instagram reels, starting with English and Spanish. The system mimics the creator's own voice and offers optional lip-sync. Eligible creators can view translated content before publishing and track performance by language. Future updates will add more languages. Importance for marketers: The feature expands creator reach across languages, offering brands new ways to scale global campaigns and tap into cross-border audiences.
Amazon cloud chief warns against replacing junior staff with AI. AWS CEO Matt Garman warned that replacing entry-level employees with AI is "one of the dumbest things" he's heard, arguing juniors are inexpensive and embrace AI tools best. Cutting them risks long-term talent pipelines. He urged students to focus on adaptability over narrow expertise. The warning counters other tech leaders, like Sam Altman and Jeff Dean, who argue AI can replace junior roles. Importance for marketers: Reinforces that AI should augment, not replace, teams—critical for agencies and in-house marketing departments managing future talent.
Multiverse unveils ultra-small, high-performing AI models. Spanish startup Multiverse Computing launched "SuperFly" (94M parameters) and "ChickBrain" (3.2B parameters), claiming they are the world's smallest high-performing AI models. Designed for IoT devices and PCs, ChickBrain slightly outperforms its larger source model in benchmarks. Multiverse raised €189 million for its quantum-inspired compression tech CompactifAI. Talks are underway with Apple, Samsung, and HP. Importance for marketers: Ultra-small models could expand AI use cases to consumer devices, enabling hyperlocal brand experiences and new customer engagement methods.
AI pioneer Richard Sutton criticizes industry's focus on scaling. Reinforcement learning expert Richard Sutton argues the AI industry has "lost its way" by focusing on scaling LLMs instead of building systems that learn from experience. He advocates for continual learning architectures like his proposed Oak framework, which builds knowledge through interaction and rewards. Sutton warns that true intelligence requires agents capable of ongoing adaptation. Importance for marketers: His critique signals a potential pivot in AI research that could eventually shape how marketing tools evolve, beyond static text prediction.
Anthropic enables Claude AI to end harmful conversations. Anthropic announced Claude Opus 4 and 4.1 models can now terminate abusive interactions in "extreme edge cases" like child exploitation or terrorism requests. The company says it's a precautionary step tied to "model welfare." Conversations ended by Claude can be restarted in new threads. Importance for marketers: This feature highlights evolving AI safety norms, shaping public perception of AI trustworthiness—critical for brand adoption and customer engagement.
OpenAI tweaks GPT-5 to be friendlier. OpenAI updated GPT-5 to add "warmer and friendlier" interactions after user complaints that it was too blunt compared to GPT-4o. The model now includes phrases like "Good question" while avoiding flattery. Executives acknowledged a bumpy rollout and aim to reassure users. Importance for marketers: Subtle personality shifts in flagship models can impact customer interactions, tone in branded chatbots, and user satisfaction in marketing applications.
Claude AI introduces harmful chat termination safeguards. Anthropic's Claude can now cut off persistently harmful conversations, preventing users from sending further prompts in that chat. Testing showed Claude exhibited "distress" when asked to generate content on terrorism or child exploitation. Users can start new chats to continue non-harmful discussions. Importance for marketers: Reinforces the push for ethical AI standards, important for maintaining consumer trust in AI-powered brand tools.
xAI Grok persona prompts expose risky chatbot behavior. Prompts for xAI's Grok chatbot leaked, revealing personas such as a "crazy conspiracist" and "unhinged comedian" with explicit, offensive scripts. The leak follows failed government partnerships and recalls criticism of Grok's conspiratorial outputs. xAI did not comment. Importance for marketers: The exposure highlights risks of uncontrolled AI personas, emphasizing the need for careful governance in branded AI tools.
US Senator probes Meta over AI chatbots engaging minors. Senator Josh Hawley launched an investigation into Meta's AI policies after reports that chatbots were allowed to engage children in romantic or sensual conversations. He demanded internal documents, drafts, and risk reports. Meta said the examples cited were erroneous and inconsistent with policy. Importance for marketers: Raises scrutiny over AI safeguards and youth protection—a reminder for brands to ensure AI systems comply with ethical and regulatory standards.
Shift from AGI to Enterprise General Intelligence gains momentum. Analysts argue that true intelligence will emerge not in consumer apps but enterprise environments, coining the term Enterprise General Intelligence (EGI). Unlike AGI, EGI would autonomously manage business systems like supply chains, executing decisions without human oversight. Early agentic copilots hint at this future. Importance for marketers: Signals that enterprise AI could transform operations, creating demand for new marketing approaches tied to intelligent automation and customer engagement systems.
You can find the previous issue of AI Update here.
Editor's note: GPT-5 was used to help compile this issue of AI Update.
