AI & Tech · Week 43, 2026
AI/Tech this week: Meta hit with $25B privacy penalty, OpenAI models land on Google Cloud, Gmail data feeds ads
A record state privacy penalty against Meta, a surprising cloud partnership between rivals, and a quiet change to what Google can do with your Gmail. Here are the five stories worth two minutes.
- 1Policy
Meta ordered to pay $25 billion in New Mexico over Cambridge Analytica data case
A New Mexico judge ordered Meta to pay $25 billion after a Santa Fe jury found it misled state residents about how it collects and protects user data. The jury found more than 43 million violations of state law. Meta's chief legal officer said the company expects to win on appeal.
Why it matters: It is one of the largest privacy penalties ever imposed on a tech company, and a sign that US states will pursue data cases on their own.
Source: cnbc.com
- 2Models
OpenAI's models are coming to Google Cloud customers
OpenAI said its frontier models and Codex coding workflows will be available to Google Cloud customers through Gemini's Model Garden and the Google Cloud Marketplace. That puts OpenAI's models alongside Google's own Gemini models in Google's catalogue.
Why it matters: Two of the biggest rivals in AI are now selling through the same shop, making it easier for businesses to mix models rather than pick one side.
Source: openai.com
- 3Policy
Google updates privacy policy so Gmail and Photos data can feed personalised ads
As of 16 October, Google's privacy policy adds an exception to its promise not to use Drive, Gmail or Photos for personalised ads: it applies unless you let those services share data with other Google services. People who connect these apps to Search's Personal Intelligence features may now see ads based on that content.
Why it matters: A privacy promise many people relied on now has an opt-in exception, so it is worth checking your Search personalisation settings.
Source: theverge.com
- 4Tools
DeepSeek open-sources DeepInfer, the software that runs its models in production
DeepSeek released DeepInfer, the inference framework behind its large-scale production services, under the Apache 2.0 licence. It includes optimised GPU kernels, asynchronous batch scheduling, structured outputs and options for deterministic, batch-invariant results. DeepSeek says it was built primarily on vLLM, with inspiration from SGLang and TensorRT-LLM.
Why it matters: Serving software, not just model weights, is now being given away, lowering the cost of running large AI models outside the big clouds.
Source: github.com
- 5Other
eMudhra quits the global website-certificate business after security incident
Indian digital trust firm eMudhra said its board voted to withdraw from the global SSL/TLS certificate business following a recent incident in which it revoked affected certificate authorities within 14 hours. The company said the line accounted for about 1% of global revenue and that its other signing and identity products are unaffected.
Why it matters: Certificate authorities underpin the padlock in your browser, and losing trust in one is usually fatal to that part of the business.
Source: emudhra.com