In partnership with
🖱
The U.S. government just kicked DeepSeek out of federal systems.
Every civilian agency has to delete the Chinese AI company's tools and can't buy anything from them again.
This isn't a suggestion it's a direct order from Homeland Security.
But wait, there's more:
Cloudflare built an AI that extracts data from messy documents in one shot
DeepSeek open-sourced its inference engine to cut AI costs by up to 75%
I’m Alex. Welcome to L8R by Innov8.
Let’s dive deep.
In today’s post:
U.S. bans DeepSeek from federal systems
Cloudflare's Clairic extracts data without the fluff
DeepSeek open-sources its inference engine
Uncle Sam Slams the Door on DeepSeek
The U.S. government just kicked DeepSeek out of its computer networks.
Officials are terrified that this cheap and powerful Chinese AI could spy on federal agencies or plant malware.
It is a massive blow to the startup that recently shook up the global AI race.
The details:
The Department of Homeland Security ordered all federal civilian agencies to rip out DeepSeek products and cancel new contracts.
The ban explicitly covers any third-party apps or tools built using DeepSeek's AI models.
Officials warned that China's national security laws could force the company to hand over sensitive U.S. government data.
Private citizens and regular businesses can still use the AI tool, as the ban only applies to federal agencies.
This federal order follows similar bans by U.S. states like Texas and New York, plus countries like India and Australia.
Why it matters:
AI models are now treated like dangerous supply-chain risks.
DeepSeek shocked the world in early 2025 by building a super-smart AI for pennies compared to Silicon Valley.
Now, the U.S. is fighting back with strict rules to protect its data. This sets a huge precedent for how governments judge foreign software.
💡 L8R's Take:
The U.S. is clearly terrified of losing its AI crown to a cheaper competitor.
Banning DeepSeek in government systems makes total sense for security, but it also protects American tech giants from real competition.
You cannot just ban your way to better innovation.
From Our Partner
See what's really happening in every deal
Aligned shows you what buyers are doing between meetings inside your most important deals then surfaces the risks, openings, and next steps. No more guessing where a deal stands or finding out a champion went quiet too late. Just real visibility, so you can act before it's too late.
Cloudflare's New AI Skips the Chat to Extract Data Fast
Normal AI models waste time acting like chatty humans when you just want raw data.
Cloudflare just fixed this with Clairic, a new model built purely to rip facts and numbers out of messy documents.
It skips text generation completely to give you structured answers in a single pass.
The details:
Clairic costs just $0.10 per million input tokens and charges absolutely zero for output.
It processes text and up to four images at once using a 65,536 token context window.
Instead of writing sentences, it spits out exact data types like dates, numbers, or lists.
Every answer includes a confidence score so you know if the AI is guessing.
In a test on product claims, Clairic was three times faster and half the cost of standard LLMs.
Why it matters:
Getting standard AI to output perfect data is a massive headache.
They often make things up or break the formatting. Clairic solves this by acting like a smart filter instead of a writer.
This makes building automated workflows for things like invoices and receipts much cheaper and far more reliable.
💡 L8R's Take:
Using a massive chatbot AI to read a simple receipt is like using a sledgehammer to crack a nut.
Cloudflare is smart to build cheap, specialized tools for boring business tasks.
Expect every major cloud provider to copy this exact approach by the end of the year.
From Our Partners
Your SOC 2, handled end to end with AI agents.
Enterprise buyers won’t put your product near their customer data without a SOC 2 report. Twenty-one state privacy laws now hold them accountable for the vendors they share data with, so their diligence lands on you.
Sprinto gets you audit ready in 14 days, across three working sessions. AI agents connect your stack, collect the evidence auditors ask for, and close the gaps as they appear. You approve, they execute.
Sprinto also answers the security questionnaires your buyers send, using the same evidence base, so security reviews stop holding up deals.
No compliance hire. No consultant. Your auditor signs off.
Anthropic's Next AI Wants to Read Your Mind
Running massive AI models just got a whole lot cheaper.
DeepSeek just dropped DeepInfer, the exact engine they use to keep their own costs rock-bottom.
They are literally letting anyone copy their incredibly efficient homework.
The details:
DeepInfer is an open-source inference engine that splits AI workloads across multiple chips to save money and power.
It currently runs the DeepSeek-V4.1-Flash model and works on both Nvidia SM100 GPUs and Huawei Ascend 950 NPUs.
The code is free for anyone to use right now on GitHub under the Apache 2.0 license.
It is not a full plug-and-play service, meaning builders still have to add their own memory cache and load balancers.
This massive release comes right after the US government banned federal agencies from using DeepSeek products.
Why it matters:
DeepSeek is famous for crushing AI costs.
Now we see exactly how they do it. Giving day-zero support to Huawei chips is a massive deal.
It proves Chinese hardware can go toe-to-toe with Nvidia. Cloud providers can use this code right now to slash their own bills.
💡 L8R's Take:
The US bans them, so DeepSeek open-sources their best tech for the whole world to use. It is a total power move.
Nvidia should be sweating because DeepSeek just proved you do not need their expensive chips to run world-class AI.
🚀 Quick L8R Summary
U.S. bans DeepSeek: Federal agencies must remove all DeepSeek AI tools and can't buy new ones.
Cloudflare's Clairic: New AI model extracts data from documents without the token-by-token dance.
DeepSeek's DeepInfer: Open-source framework that cuts AI inference costs and energy use.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋