Apple Just Stopped Building Its Own AI Foundation. Your Vendor Map Changed.
Apple and Google entered a multi-year collaboration under which the next generation of Apple Foundation Models will be based on Google’s Gemini models and cloud technology. The partnership is not exclusive, Apple still works with OpenAI, and Apple Intelligence will continue to run on-device and through Private Cloud Compute. Previous reports put Apple’s annual payment at around $1 billion.
The split Apple chose is the right one: borrow the best model, run it on infrastructure you own. The failure I see most often is the inversion: proprietary models, unaudited vendor infrastructure. Before your next model contract renewal, ask what your own version of Private Cloud Compute looks like. For organizations in the EU already mapping vendor dependencies under the AI Act, this is the moment to do it — not after a supply chain shift forces your hand.
Read more: Google Blog | TechCrunch | CNBC
AI Video Just Passed the Turing Test. Your Trust Protocols Didn’t.
In a study of 1,043 participants, over 90% could not reliably distinguish Runway’s Gen-4.5 outputs from real video. Overall accuracy was 57.1% — barely above chance. For animals and architecture, detection fell below chance: generated content fooled participants more reliably than real footage did.
Trust verification designed around human judgment has a hard expiry date. The Runway data suggests it has already passed. The dangerous failure mode isn’t external manipulation — it’s internal: AI-generated outputs re-entering workflows as trusted inputs. Your governance framework needs to shift from detection to provenance. Not “is this real?” but “where did this originate and what chain of custody exists?” For EMEA organizations, C2PA provenance standards will likely become an AI Act compliance expectation. Start treating that as a process design problem now.
Read more: Runway Research | TechRadar
OpenAI Monetizes the Mass Market. What That Means for Your AI Procurement.
OpenAI launched ChatGPT Go globally at $8 per month — its fastest growing plan. Ads will be tested in the US for free and Go tier users, while Plus, Pro, Business, and Enterprise remain ad-free. OpenAI projects a $17 billion cash burn in 2026 with 800 million monthly users, only 5% of whom pay for anything.
The tool sprawl pattern is always the same: a department starts on the free tier, usage grows, and six months later the organization runs an undocumented AI portfolio nobody authorized. If your employees use ad-supported AI for work tasks, your data boundary assumptions are already under pressure. The answer to “what is our AI procurement strategy?” cannot be “our people use the free version.” The ad-free positioning of Claude and Gemini creates temporary leverage in vendor negotiations. Use it while it lasts.
Read more: OpenAI | TechCrunch | CNN
Worth Watching: ElevenLabs and the First Rights-Cleared AI Content Template
ElevenLabs launched The Eleven Album with 13 artists representing over one billion streams. Each artist retains full ownership and collects all streaming revenue. The model underneath — licensed training data, explicit consent, revenue-sharing with Kobalt Music and Merlin — is the first large-scale demonstration of what a compliant AI content ecosystem looks like. For any organization exploring AI-generated media, this is the rights and licensing template to benchmark against before EU AI Act provenance obligations sharpen.
Read more: ElevenLabs Blog | Variety
Leave A Comment