Saazora
SaaS News
Nov 13, 2026 16 mins read 395 views

AI & SaaS News, Nov 13: Gemini 4 Hydrogen, Outlook Outage

Gemini 4 Hydrogen pricing and access, the Gemini Decisions API, Claude's new usage policy, OpenAI Daybreak, Gemini on CarPlay, the Microsoft 365 outage, Nvidia's interim CEO and a lawsuit to freeze AI models.

Quan Le
Quan Le
AI Tools Reviewer
AI & SaaS News, Nov 13: Gemini 4 Hydrogen, Outlook Outage

Last checked November 13, 2026 · News roundup

Today in brief: Google replaced its Flash line with Gemini 4 Hydrogen, a faster model priced at $1.50 per million input tokens that free users can reach through auto model selection. Microsoft 365 spent Thursday afternoon recovering from an outage that took Outlook and Teams offline for hours, and Nvidia named CFO Colette Kress interim CEO after Jensen Huang resigned. Four nonprofits also asked a San Francisco court to stop five AI labs from releasing new models.

Part 1 AI

Gemini 4 Hydrogen price: $1.50 in, $7.50 out, and free users get it automatically

Announced November 12 · Affects Gemini app users, Search users and API developers

Google launched Gemini 4 Hydrogen on November 12 as its new everyday model, and according to 9to5Google it retires the "Flash" name. On the Gemini API it costs $1.50 per million input tokens and $7.50 per million output tokens, flat across the full 1-million-token context window. That is 25% below the introductory rate of Gemini 4 Argon and Gemini 4 Carbon ($2 and $10), and well under the $4 and $20 those two move to later.

Google says Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default "high" thinking level, one point above Argon, and 54 at the new "xhigh" level. It also claims about 2.5 times faster task completion than Gemini 3.8 Flash.

Who gets it: developers through the Gemini API, AI Studio (all tiers) and Android Studio; Google AI Pro and Ultra subscribers in Antigravity and, first, in Gemini for Google Sheets; everyone in Google Search; and Gemini app users through the model picker. Free app users reach it only when auto model selection decides a prompt needs it.

What to do: if you run high-volume agent or coding jobs on Argon, test Hydrogen on the same prompts before November 30, when Carbon's intro price ends. Our Gemini 4 Argon price and access guide has the worked cost examples for the bigger model.

Gemini API list prices per million tokens, November 2026

Input Output

Decisions API (gemini-decision-1)

$0.10

$0

Gemini 4 Hydrogen

$1.50

$7.50

Argon and Carbon, intro rate

$2

$10

Argon and Carbon, standard rate

$4

$20

Sources: Google blog posts on Gemini 4 Hydrogen (Nov 12, 2026) and the Decisions API (Nov 11, 2026); 9to5Google on Carbon's intro price running through November 30. Bars are scaled to the $20 maximum. Cached-input discounts are not shown.

What is the Gemini Decisions API? Up to 1,000 yes/no calls for $0.10 per million tokens

Announced November 11 · Affects developers building agents and classifiers

The Decisions API is a new Gemini endpoint that returns choices instead of text. You send one input (text, images, audio, video or PDFs, up to 256,000 tokens) plus up to 1,000 questions, and the gemini-decision-1 model answers each as a yes/no probability, a pick from 2 to 255 options, or a score on a rubric, with a confidence value from 0 to 1.

The pricing is unusual: $0.10 per million input tokens and nothing for output. Google bills the shared input once per request, so checking a 10,000-token document costs about a tenth of a cent for that input, plus each question's own instruction tokens. Google puts end-to-end latency at 150 to 250 milliseconds for inputs up to 16,000 tokens, and says it is nearly five times cheaper than Gemini 3.5 Flash-Lite on its test suite.

Who it is for: teams that use a full chat model today just to route requests, tag tickets, run policy checks or grade retrieved context. It is live in the Gemini API and AI Studio, with a free tier for prototyping; Google Cloud support is "coming soon." Swap one routing step first and compare cost and accuracy before moving more.

Claude usage policy changes took effect November 12: what is now banned

In force November 12 · Affects every Claude user, free plan included

Anthropic's yearly Usage Policy update, published October 8, applies from November 12. The headline change bans "sustained and needless" abusive or cruel behavior toward Claude. Anthropic says it targets only extreme, repeated cruelty with no purpose; ordinary frustration, dark fiction and research testing are not covered, and the main enforcement remains Claude ending the conversation.

The other changes matter more for businesses. Tracking people without consent, or using Claude to decide who should be investigated, arrested or charged, is explicitly prohibited. A new section groups fake-account networks and influence operations under one rule. The weapons ban now names software and drone-arming systems. The blanket ban on personalized campaign targeting is gone, though deceptive targeting is still banned. If you connect Claude to hardware that acts on its own, a human operator must be able to stop it.

Separately, Claude Meetings started rolling out to Team and Enterprise plans on Google Meet on November 12, according to Unite.AI. Notes, transcripts and in-call answers are free during the beta. Before installing third-party add-ons, see how to vet a Claude skill.

OpenAI Daybreak Defense Factory and Red Team: AI agents that patch and pen-test code

Announced November 12 · Affects security and engineering teams

At its "Cyber at the Frontier" event on November 12, OpenAI added two products to its Daybreak security platform. Daybreak Defense Factory runs agents that scan a company's codebases, confirm vulnerabilities and draft fixes; per The Register, the agents currently run on GPT-6 Cyber and hand engineers a merge-ready pull request with a test showing the patch works. A person still reviews and approves every change before it ships.

Daybreak Red Team points a fleet of agents at your web applications, using source code and app context to probe attack paths. It can also retest findings from scanners, bug bounties and disclosure reports after a fix lands. Red Team is in early access, and OpenAI vets Daybreak applicants and their intended use.

What it costs: OpenAI has not published pricing for either product. What to do: if you already pay for a code scanner, ask your OpenAI account team whether Defense Factory can ingest its findings, since triage of existing alerts is where most teams lose time. Do not point Red Team at systems you do not own or have written permission to test.

How to use Gemini on Apple CarPlay

Rolled out November 11 · Affects iPhone drivers

Gemini now has its own icon on the Apple CarPlay screen, so iPhone owners can start a hands-free Gemini Live conversation without going through Google Maps, which was the only route before. Google pitches it for rehearsing a presentation, catching up on email, hearing your Daily Brief or finding nearby places, and every chat syncs to your Gemini history on the phone.

To turn it on: update the Gemini app on your iPhone, allow microphone access, and use Gemini Live at least once on the phone. The icon then appears on the CarPlay display. It works in every language Gemini Live supports. Engadget's report lists no paid-plan requirement, and Google did not mention one. Google says Personal Intelligence is already on Android Auto and is coming to cars with Google built-in.

Part 2 SaaS & business

Is Microsoft 365 down? What happened to Outlook and Teams on November 12

November 12 · Affects Microsoft 365 business and school tenants

A Microsoft 365 outage took down Outlook mail access for much of Thursday afternoon in the US. Downdetector logged 138,058 Outlook problem reports at the peak around 1 p.m. ET, according to Reuters. Microsoft tracked it as incidents EX1550543 and MO1550821 and listed Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot and Defender XDR among the affected services, plus Intune and Purview.

Microsoft's explanation: servers failed automated health checks and dropped out, pushing load onto the rest, and a core process on those servers became overwhelmed. It also reviewed a few recent configuration updates but had not confirmed a root cause. Recovery was gradual; at 6 p.m. ET Microsoft was still bringing servers back. The ripple reached AI tools too: Reuters reported OpenAI and Anthropic saw problems with their Microsoft connectors.

What to do: follow MO1550821 in the admin center's Service health page rather than rebuilding Outlook profiles. Once mail is back, check Intune check-ins and Defender alerts, and log the hours lost in case your agreement includes service credits. If the outage has you thinking about a second mail provider, our Hostinger Email vs Google Workspace comparison covers the costs.

Microsoft 365 outage, November 12 (Eastern Time)

12:30 pm

User reports for Outlook start climbing

~1:00 pm

Outlook reports peak at 138,058 on Downdetector

1:10 pm

Microsoft flags Exchange Online mailbox connection problems

2:16 pm

Deployments and configuration changes paused

4:01 pm

A new health-check method returns a subset of servers

5:11 pm

Microsoft says fixes have not brought the expected relief

6:05 pm

More servers return and "remain stable so far"

Sources: TechRadar live blog and Reuters via The Korea Times (Nov 12–13, 2026); Microsoft status updates as logged by WindowsForum in UTC and converted to ET (UTC−5). Times are approximate.

Who is Nvidia's CEO now? Colette Kress steps in after Jensen Huang resigns

November 12 · Affects Nvidia investors and AI infrastructure buyers

Jensen Huang resigned on November 12 as Nvidia's president, CEO and board member, effective immediately, according to the company's SEC filing. CFO Colette Kress is interim CEO and keeps the CFO job, and Harvey C. Jones becomes the board's first chair. The board is searching inside and outside the company for a permanent CEO.

Nvidia gave no reason. The resignation follows a week of reports about Huang's personal relationships and an independent board investigation, according to Decrypt. Shares fell 7.7% to $160.28 shortly after the open and closed down more than 4%, CNBC reported. Huang remains the largest shareholder with about 3.6%.

For businesses buying GPUs or AI cloud capacity, nothing changed on products, prices or delivery dates in the announcement. The next checkpoint is Nvidia's fiscal third-quarter report on November 17, which the company says is still on schedule. Watch the outlook it gives on supply, since GPU costs flow into what AI and SaaS tools charge you.

Nvidia leadership: before and after November 12

RoleBeforeAfter
CEOJensen HuangColette Kress (interim)
CFOColette KressColette Kress (keeps the role)
Board chairNo separate chair roleHarvey C. Jones, the first ever
Huang on the boardDirectorResigned
Q3 FY2027 resultsNovember 17November 17 (unchanged)
Source: Nvidia Form 8-K exhibit filed with the SEC, November 12, 2026.

Lawsuit to freeze AI: nonprofits sue OpenAI, Google, Anthropic, Meta and SpaceXSI

Filed November 12 · Affects AI labs, and in theory every user of their models

Four nonprofits (the Irreplaceable Education Fund, Public Citizen, Heat Initiative and the Alliance for Secure AI) sued Anthropic, Google, Meta, OpenAI and SpaceXSI in San Francisco Superior Court on November 12, the Associated Press reported. They want a judge to stop the companies from developing or deploying new AI models.

The case rests on California's unfair competition law. The groups argue the labs cannot prevent "catastrophic" harm, citing documented AI-assisted cyberattacks and the companies' own statements that the technology could pose existential risks. They also claim the systems interfere with the use of property such as computer systems. None of the companies had commented when the AP published.

What changes for you today: nothing. ChatGPT, Gemini, Claude and Meta AI keep working, and no hearing date was reported. The filing lands as Anthropic heads toward its IPO and as Washington debates new AI rules, so watch whether a judge agrees to hear an early request to pause releases.

Also on the radar

  • Databricks Genie: an always-on version of the data agent now runs multistep jobs unsupervised, with spending limits set through Unity Gateway (TechCrunch, Nov 11).
  • TSMC: its board approved up to $1 billion for AI labs, split between $500 million of OpenAI preferred stock and $500 million of Anthropic shares at the IPO price (Ars Technica citing Reuters, Nov 11).
  • Norwest: raised a $4 billion fund, its largest, aimed at AI apps, voice AI and agentic cybersecurity (TechCrunch, Nov 12).
  • Fractile: the AI inference chip startup raised a $720 million Series C at a $7 billion valuation (Tech Startups citing Reuters, Nov 12).

Your Friday checklist from today's news

Microsoft 365 admins

☐ Confirm MO1550821 shows as resolved in Service health

☐ Review Intune check-ins and Defender alerts from Thursday

☐ Write down outage hours for any credit claim

Gemini API developers

☐ Rerun one Argon workload on Hydrogen and compare cost

☐ Move a routing or tagging step to the Decisions API

☐ Note November 30, when Carbon's intro price ends

Claude Team and Enterprise admins

☐ Decide whether Claude may join Google Meet calls

☐ Recheck any monitoring use case against the new surveillance rules

Security teams

☐ Apply for Daybreak Red Team early access if you test web apps

☐ Ask how Defense Factory handles your existing scanner findings

Built by Saazora from the vendor announcements cited in this roundup, November 13, 2026.

Friday questions, answered

Is Gemini 4 Hydrogen free to use?

Partly. Free Gemini app users get Hydrogen when auto model selection picks it for a prompt, Google Search uses it for everyone, and AI Studio offers it on all tiers. Antigravity access needs Google AI Pro or Ultra. Paid API use costs $1.50 per million input tokens and $7.50 per million output tokens.

Is Outlook working again after the November 12 outage?

Microsoft was restoring servers gradually on Thursday evening, and by 6 p.m. ET reports had fallen well below the afternoon peak. Check incident MO1550821 in your Microsoft 365 admin center for the official status. Microsoft has not yet published a final root cause.

Who replaced Jensen Huang as Nvidia CEO?

Colette Kress, Nvidia's chief financial officer since 2013, is interim CEO and keeps her CFO role. The board is considering internal and external candidates for the permanent job. Harvey C. Jones is the company's first board chair, and Nvidia still plans to report earnings on November 17.

Can Anthropic ban you for being rude to Claude?

Only for extreme cases. The policy in force since November 12 covers sustained, needless cruelty, not frustration, blunt feedback or dark creative writing. The main response is Claude ending that chat, and you can start a new one right away. Anthropic's general terms still allow stronger action for policy violations.

Sources: Google on Gemini 4 Hydrogen (Nov 12, 2026) · 9to5Google (Nov 12, 2026) · Google on the Decisions API (Nov 11, 2026) · Anthropic Usage Policy update (Oct 8, 2026; effective Nov 12) · Unite.AI on Claude Meetings (Nov 12, 2026) · The Register on Daybreak (Nov 12, 2026) · Engadget on Gemini for CarPlay (Nov 11, 2026) · TechRadar live blog (Nov 12, 2026) · Reuters via The Korea Times (Nov 13, 2026) · WindowsForum incident log (Nov 12, 2026) · Nvidia 8-K exhibit (Nov 12, 2026) · Decrypt (Nov 12, 2026) · CNBC (Nov 12, 2026) · Associated Press via ABC News (Nov 12, 2026) · TechCrunch on Databricks (Nov 11, 2026) · Ars Technica on TSMC (Nov 11, 2026) · TechCrunch on Norwest (Nov 12, 2026) · Tech Startups (Nov 12, 2026)

Tags: #AI news today November 13
Quan Le
ABOUT THE AUTHOR

Quan Le

Quan Le is the founder and editor of Saazora, an independent site for news and reviews of SaaS and AI tools. He has worked in affiliate publishing since 2015, writing product reviews, verifying deals and running SEO for review sites. On Saazora he focuses on software for small businesses, creators and marketing teams, and evaluates tools by how well they fit real workflows: onboarding, integrations, pricing and long-term value, not the length of their feature list.

Read Quan's full profile · LinkedIn · contact@saazora.com

KEEP READING

Related Reviews