Gemini 4 Carbon leads today's AI news as ChatGPT adds API credits

Google's latest model announcement and OpenAI's new subscriber benefit raise two practical questions: who can use the technology, and what can they do with it? The October 29 edition of AI World Today: Daily News also examines the cost of expanding cloud infrastructure, Nvidia's proposed Hugging Face deal, and an AI partnership at the Rijksmuseum.

Listen to AI World Today: Daily News on Spotify, Apple Podcasts or Amazon Music. Find more episodes in our podcast archive.

In its October 28 announcement, Google positions Gemini 4 Carbon for sustained engineering work and software security. Initial access begins with a subset of cyber defenders, followed by a gradual expansion through Fairwind. A wider release date has not been announced.

The Fairwind Program prioritizes defensive work at organizations including governments, healthcare providers and telecommunications services. Its purpose is to give defenders an opportunity to strengthen systems as models become more capable.

Our view: the most useful evidence will come from reviewed changes to real software. An engineering assistant needs to investigate a problem, explain a proposed repair and help establish that it works. Google's performance claims make that a promising area to watch, while broader testing will determine how dependable the workflow becomes.

OpenAI's October 28 release notes introduce API credits for eligible paid Pro and Business subscribers.

The eligibility terms specify full-period monthly Pro allowances of $100, $200 or $500, matching the eligible tier. For Business, paid Standard seats contribute $20 and Premium seats contribute $100, pooled to a maximum of $500 per workspace per month. Annually prepaid Business workspaces receive an upfront annual grant, capped at $6,000.

Activation requires an eligible paid subscription active for at least seven days and an owner linking an API organization they own. Country and rollout availability also matter.

The developer guide explains that the first grant is prorated and credits expire at the end of the applicable billing cycle. They fund supported API activity without increasing ChatGPT or signed-in Codex limits. Codex usage through API keys is excluded. Plus subscribers do not qualify, and the promotional benefit can change.

For an eligible small team, this creates space to test a narrowly defined idea. A useful pilot should answer a concrete question, such as whether an assistant can reduce the time spent sorting incoming requests.

Alphabet's results and Microsoft's results, both released October 28, cover the quarter ended September 30.

Microsoft's figure is calculated as $58.000 billion in operating cash flow minus $39.367 billion in cash additions to property and equipment. Alphabet reports −$9.065 billion using the same subtraction.

The cloud categories differ. Alphabet also says TPU systems sales contributed meaningfully to growth, so its Cloud result should not be treated as recurring AI subscription revenue.

Alphabet remained profitable despite negative free cash flow. Infrastructure purchases can consume cash immediately while their accounting cost is spread over years. These company-wide results do not isolate either company's AI profitability.

Our reading is that capacity, utilization and lasting returns deserve attention alongside headline growth. Revenue alone cannot show whether an infrastructure investment will pay off.

Reuters reports, citing two sources, that the Justice Department is examining the deal's competitive effects. The report establishes neither wrongdoing nor a decision to block it.

Nvidia announced the proposed roughly $13 billion transaction in September. Its SEC filing anticipates completion in the first half of 2027, subject to conditions including regulatory approvals.

Nvidia has committed to supporting other hardware, clouds and models on Hugging Face. The editorial question is how that choice would work in practice after completion. For developers, the tools and services surrounding an open model can matter as much as the model's availability.

The Rijksmuseum's October 27 announcement sets out a three-year AI partnership with Prosus. Specialists will work full time on museum projects, aiming to help more people discover and experience the collection.

Prosus describes an initial assessment of goals, data and workflows before developing applications. A named new visitor product and release date have not been announced.

The museum already introduced visual search in December 2025, allowing exploration by similarities in shape, colour, style and object type, as its 2025 review explains. That feature predates the partnership.

For someone who recognizes a painted detail but cannot name its artist or period, this points to an approachable way into art. The partnership's eventual value will depend on how well its applications turn that kind of curiosity into discovery.

Follow AI World Today: Daily News for our next episode. The podcast is narrated by an AI voice.