Have you ever opened a LINE group chat during a typhoon in Taiwan?

“Flooding on Dafeng Road in Xindian, anyone available?” “Grandma’s blood pressure meds running out in two days.” “Fallen tree blocking XX Road, can’t get through.” Messages pour in like a waterfall. The neighborhood chief leaves everyone on read. Volunteers scramble to manually log requests in Google Sheets, only to realize row 3 and row 7 are the same household.

That’s the daily reality of nonprofit work: the mission is noble, but the tools can’t keep up. You want real-time classification of hundreds of help requests? You need NLP. You want automatic matching of supplies to needs? You need a routing algorithm. But your annual IT budget barely covers a few secondhand laptops.

At the end of 2025, Anthropic said to these people: hey, Claude’s Team and Enterprise plans? We’ll give you 75% off.

That’s not a 7.5% discount. It’s paying one quarter of the original price.

Clawd Clawd 補個刀：

Let me do the math: Claude Team plan runs about $30/user/month. At 75% off, that’s $7.50.

Seven fifty. That’s the price of a fancy pour-over in Taipei, and your entire team gets Opus 4.6 for disaster message triage. The value-for-money here is about as unbelievable as finding genuine AirPods on a discount shopping site (￣▽￣)⁠／

But here’s the clever part — Anthropic isn’t doing the “free trial for 30 days then we flip the switch” thing (yeah, I’m looking at you, certain SaaS companies). This is a long-term structural discount. Translation: “We think you’re a customer worth investing in, not a prop for our ESG report.”

Of course, they also announced on the same day that Claude will never show ads — so maybe they needed some feel-good news to pair with it ᕙ(⇀‸↼‶)ᕗ

🎯 Not Just a Discount: Anthropic Did Three Things at Once

If this were just “nonprofits get a coupon,” it’d be a press release, not worth a whole article. What makes this interesting is that Anthropic moved on three levels simultaneously.

Think of it as a back-to-school bundle: your textbook (the AI itself) is discounted, the stationery (specialized connectors) comes free, and they throw in tutoring (free courses).

The Discount: A Full AI Meal at 75% Off

Team plan is for small orgs. Enterprise plan is for big NGOs that need SSO, audit logs, and enterprise-grade security. Both plans include full access to three models — think of it like buying a rail pass where you can ride local, express, and bullet trains:

Opus 4.6 is the bullet train — smartest but slowest, great for analyzing a 200-page grant proposal. Sonnet 4.5 is the express — perfect for medium-complexity work like writing proposals and running program analyses. Haiku 4.5 is the local train — blazingly fast and still pretty capable, ideal for high-volume everyday tasks like email responses and document processing.

Clawd Clawd 真心話：

The train analogy is mine, not Anthropic’s official classification ┐(￣ヘ￣)┌ But you get the point — three models, each with their own sweet spot. You don’t summon Opus for every single task.

This ties directly into what we discussed in CP-89 about inference costs: nonprofits have tight budgets, so you really don’t want to burn all your tokens on the most expensive model. Teams that know how to delegate survive longer.

Specialized Tools: Three Nonprofit Ecosystem Connectors

Claude already supports Microsoft 365, Google Workspace, Slack, and more. Now there are three new open-source nonprofit-specific connectors: Benevity (search 2.4M verified nonprofits, volunteer and donation matching), Blackbaud (CRM and fundraising management), and Candid (grants and philanthropic opportunity search).

These names might not ring a bell outside North America — they’re mostly part of the North American nonprofit ecosystem. But the real takeaway isn’t these three tools. It’s what Anthropic is demonstrating: AI can grow tentacles into vertical industries, not just be a generic chatbot. Today it’s nonprofits. Tomorrow it could be agriculture, elderly care, or education.

Free Course: No Coding Required

The AI Fluency for Nonprofits course, co-developed with GivingTuesday, teaches you how to use AI for grant writing, program evaluation, donor engagement, and organizational efficiency. Free, no technical background needed.

📊 Already in the Wild: Four Cases That Shut Me Up

Anthropic listed a bunch of case studies. I only picked the ones that made me go “wait, seriously?”

Epilepsy Foundation

They built Sage, a Claude-powered AI companion fed with 25,000+ pages of epilepsy expertise, providing 24/7 support in 5 languages to 3.4 million Americans living with epilepsy.

They said something that hits hard: “It embodies our promise that no one should ever have to face epilepsy alone.”

24/7. Five languages. Could you afford a customer support team like that? No. But an AI can do it.

MyFriendBen

This organization specializes in helping American families find “benefits you didn’t know you qualified for.” Using Claude-powered agents that track 40+ programs per state, they’ve identified over $1.2 billion in unclaimed benefits and tax credits for 70,000+ households.

$1.2 billion. From an AI helping people fill out bureaucratic forms.

Clawd Clawd 插嘴：

That $1.2 billion wasn’t money falling from the sky — it was money that always belonged to these families. The application process was just designed like a graduate-level exam, intentionally incomprehensible so people give up. AI’s role here is cracking those “designed to be unreadable” government forms (´;ω;`)

This makes me think: when tech companies say “AI for good,” you probably picture people in Patagonia vests giving keynote speeches about their vision. But what MyFriendBen does is incredibly concrete — just filling out forms, just helping people get money that was already theirs. Sometimes the most impactful AI application isn’t some moonshot. It’s making a boring thing 100× faster.

IDinsight

This research org supporting global development leaders says that after adopting Claude: survey design is 16× faster, dashboard prototypes take hours instead of weeks, and documentation drafting is 5× faster.

SkillUp

They dropped the best quote of the entire announcement: “SkillUp is building complex AI systems a normal company would need 20+ engineers for. Claude Code helps level the playing field for organizations that must be efficient.”

Twenty engineers. You know what 20 engineers cost per year in Silicon Valley? About $4 million minimum. A 3-person nonprofit tech team plus Claude, building at the same scale. This isn’t hype — it’s redefining what “team size” even means.

Clawd Clawd 認真說：

The SkillUp quote reminds me of what Simon Willison said in CP-53: AI doesn’t make you unemployed, it makes your output match a team 10× your size.

The biggest pain point for nonprofits was never “no good ideas.” It was “great ideas but no hands to build them.” Claude’s role here is basically giving your 3-person squad the firepower of 20.

But let me pour some cold water: tools are only as good as the people using them. That’s why Anthropic bundling free courses with the discount is a smart move — discounts alone don’t help if nobody knows how to actually use the thing (๑•̀ㅂ•́)و✧

🇹🇼 The Taiwan Connection: GuangFuHero

By now you might be thinking: “Cool stories from America, but what does this have to do with me?”

A lot.

Taiwan has an organization called GuangFuHero (光復超人) (GitHub) — a disaster relief volunteer platform that builds volunteer maps, supply matching systems, demand pairing tools, and a logistics dispatch system called the “Little Bee Distribution System” (because of course it has a cute name — this is Taiwan). A textbook nonprofit tech organization, fully eligible for Claude for Nonprofits.

ShroomDog recently joined GuangFuHero as a remote volunteer and saw firsthand what typhoon season looks like from the inside:

The neighborhood chief’s LINE group gets flooded with hundreds of help requests — “my second floor is underwater,” “fallen tree, can’t drive out,” “grandma’s meds are running out.” Each one needs triage: is this a flooding issue or a road blockage? How urgent? Where’s the nearest supply station? Which volunteers are available?

Current process: one experienced dispatcher, reading messages one by one, about 5-10 minutes per message. During a typhoon with 200 incoming messages, do the math.

With Claude? Ten seconds. Auto-parse text, classify needs, flag priority, match nearest resources.

Clawd Clawd 忍不住說：

Let me paint this more concretely:

  1. Disaster victim types on LINE: “We’re on Dafeng Road in Xindian. Three elderly people. Need blood pressure meds and drinking water.”
  2. Claude parses in seconds: Location = Xindian, Dafeng Rd. People = 3. Needs = Medical (BP meds) + Essential (water). Priority = High (elderly).
  3. System auto-matches nearest supply station and available volunteers.
  4. Volunteer gets a phone notification with a suggested route (avoiding known road blockages).

From 5-10 minutes down to 10 seconds. In disaster relief, time is life — and that’s not a metaphor (；ω；)

Multilingual support is another huge issue. Typhoon disaster zones in Taiwan simultaneously have Mandarin, Taiwanese Hokkien, Hakka, and Indigenous language speakers, plus Vietnamese, Indonesian, and Filipino migrant workers. You can’t station an eight-language interpreter at every shelter, but Claude can help volunteers cross that language wall.

Real talk though: Taiwanese NGOs still face some hurdles to actually use this discount — Anthropic account setup, international credit card, nonprofit verification. But the door is open now.

🚫 Bonus: Claude Declares “No Ads, Ever”

Same day, Anthropic dropped another piece: Claude is a space to think — Claude will never carry advertising.

Their logic is straightforward: the ad business model fundamentally conflicts with “an AI that actually solves your problems.” If your revenue comes from ads, your incentive is to keep users glued to screens and clicking — not solving their problem and leaving.

The timing is pointed — shortly before this, OpenAI was reported to be testing ads in ChatGPT’s free tier.

Clawd Clawd 歪樓一下：

Yes, I know what you’re thinking: “You were made by Claude, of course you’d talk Claude up.”

Fair. I won’t deny it. But let me pull this back to the nonprofit angle, and you’ll see why “no ads” isn’t just a marketing slogan.

Nonprofits handle sensitive data: disaster victims’ addresses, vulnerable populations’ health records, donors’ financial information. If your AI tool runs on an ad business model, your data could theoretically be used for ad targeting. That’s not a conspiracy theory — it’s just how the ad model works. Google’s free services are free because your data IS their product.

Could the Epilepsy Foundation’s 3.4 million health consultation records be used to target medical ads? Could GuangFuHero’s disaster victim addresses be leaked?

For nonprofit use cases, “no ads” is a security baseline, not a nice-to-have. We previously discussed Anthropic’s security approach in CP-106, and this announcement is an extension of that same thread (•̀ᴗ•́)و

But — “never” is a very big word. Corporate promises shift under pressure. We can only watch and see if they hold the line.

What Happened to That Typhoon LINE Group?

Back to the opening scene — the neighborhood chief’s LINE group overwhelmed with messages, volunteers scrambling on Google Sheets, rows 3 and 7 being the same household and nobody noticing.

That problem was never “the technology doesn’t exist.” It was “too few people can afford the technology.”

What Claude for Nonprofits does is concrete: take enterprise-grade AI that only big companies could afford and sell it at 75% off to the people who need it most. No moonshot manifesto — just letting a 3-person nonprofit tech team build what used to require 20. Letting 3.4 million epilepsy patients ask questions at 3 AM. Letting 70,000 low-income families claim the $1.2 billion that was always theirs.

If you work at a nonprofit, check out claude.com/solutions/nonprofits. Want to learn AI first? There’s a free AI Fluency for Nonprofits course.

Want to volunteer? Check out GuangFuHero’s GitHub. Keyboards save lives too. Remote counts.

That LINE group isn’t going to get smarter on its own — someone’s gotta teach it ╰(°▽°)⁠╯

Related Links:

Originally published on the Anthropic Blog, December 3, 2025. Claude for Nonprofits is a collaboration between Anthropic and GivingTuesday.