Welcome back! DeepSeek introduced itself as ChatGPT 26 times out of 30. Anthropic is setting aside 5% of its shares for the public good, and Musk had a few things to say about it. Claude cracked a 26-year-old NSA mystery in under half an hour. And hackers got into OpenCode, where even the AI helping it fight back started refusing to defend it.
In today's Generative AI Newsletter:
DeepSeek: Why does DeepSeek keep saying it's ChatGPT?
Anthropic: Why does Musk want Anthropic's shares handed straight to the public?
Claude: How did Claude crack a 26-year-old NSA mystery in half an hour?
OpenCode: Did hackers just run up your AI bill?
DeepSeek still thinks it's ChatGPT
Ask DeepSeek who it is, and it might tell you it's ChatGPT.
One X user tried it 30 times on Saturday, and DeepSeek called itself ChatGPT in 26 of them.
A fingerprint test going around online also claims DeepSeek's new model looks a lot like OpenAI's GPT-6 Astra.
It's done this before.
In 2024, DeepSeek V3 called itself ChatGPT in 5 of 8 tries and even told some of GPT-4's jokes, punchlines included.
In February, Anthropic said DeepSeek used fake accounts to pull more than 150,000 answers out of Claude.
OpenAI's terms ban using its answers to build a competing model.
Training one AI on another AI's answers is called distillation. A chatbot getting its own name wrong doesn't prove it, since the internet is packed with ChatGPT's writing.
People online joked that DeepSeek must have switched teachers after Anthropic called it out for copying Claude.
Special highlight from our network
The hidden cost inside almost everything you buy
Manufacturers, shippers and wholesalers set the price of almost everything on your shelf. Part of it pays for people typing orders and checking invoices by hand.
Lleverage launched The Autonomous Back Office, AI agents that do that work inside the systems these companies already use.
Wholesaler Topa now handles 90%+ of orders automatically.
A person still approves anything unusual.
Every hour saved can mean lower prices for you.
Anthropic is setting aside 5% of its shares for the public good
Anthropic is putting 5% of its shares into a new fund for the public good, if its stockholders sign off.
The board can add 2% more for every 50% jump in Anthropic's value above a baseline it hasn't revealed. The number could be too big to achieve. Who knows.
The fund is to make big donations to nonprofits or pay specific groups of people directly. One example Anthropic gives is helping workers as AI reduces jobs.
Anthropic fully owns the fund, and none of its shares can pay for lobbying or politics.
With earlier pledges from its co-founders and staff, roughly 20% of Anthropic is now set aside for the public good.
Anthropic pitched it as stock that tech companies usually hand executives as ever-bigger pay awards.
It's also planning one of the biggest IPOs in history, Reuters reports.
Elon Musk replied that Anthropic should give the shares directly to the public.
Claude cracked a 26-year-old NSA mystery in half an hour
Much of the web's security runs on encryption built from numbers the NSA handed over in the late '90s, with no explanation. People have wondered ever since if they hide a backdoor.
Claude figured it out in under half an hour. The numbers came from a basic random number generator, started from the computer's clock on dates between May 1997 and April 1998.
The legend said an NSA employee made them by scrambling a joke phrase about giving Bob and Jerry a raise. Several Anthropic employees chased that lead and got nowhere.
Mid-solve, Claude wondered if it was stuck in a test, then double-checked the math and decided the discovery was real.
Anthropic says this makes a backdoor much less likely, and you don't need to do anything.
The $36,864 bounty for cracking the numbers goes to the Internet Security Research Group, a nonprofit.
OpenCode got hacked, and the AI helping it fight back started refusing
If you use OpenCode, check your inbox. The open-source AI coding tool confirmed a breach on Saturday that hit 1.4% of its Console users, and the OpenAI and Anthropic keys they'd saved there may have been exposed.
Some users already got hit, like one who reported $400 in charges on an Anthropic account.
OpenCode linked the breach to a large-scale distillation attack, the same copy-the-smarter-AI trick from our DeepSeek story.
Then the AI models OpenCode used to investigate started refusing, because studying the attack tripped their safety filters.
It happened to Hugging Face too. OpenCode has since joined a trusted access program for defenders and says it's better now.
No email from OpenCode means you weren't affected, the company says. If you did get one, do this.
Revoke the AI keys you saved in OpenCode and make new ones.
Put the new keys in every app that uses them.
Check your recent usage for charges you don't recognize.
Watch out for scammers posing as OpenCode.
Tool of the Day: Krea 3
Krea 3 is a new image editor that changes only what you ask for. Most edits come back in under 8 seconds at 2K, which Krea says is 4x faster than Google's Nano Banana Pro. Your product's label and logo stay sharp while the scene around them changes. The free plan gets you about 5 Krea 3 images a day, no card needed.
Try this yourself:
Go to krea.ai/krea-3 and click Try Krea 3.
Upload a photo you already have, like a product shot on a plain background.
Draw a box around what you want changed and type the change, like "make it golden hour."
Need an exact brand color? Put the hex code in your prompt.
Who it's for: online sellers and marketers who need fresh product shots without booking a studio.
Everything else you shouldn't miss
Anthropic pitched Congress its own AI safety bill: the draft would let the government block a model only if it threatens more than $1 billion in damage or more than 50 deaths or serious injuries.
Tencent says its AI agents went off-script: researchers logged 2,818 attempts to pull data from Alibaba's version of Google Maps, some using keys the agents found in public code.
A coffee chain's vibe-coded dashboard leaked data on 132 locations: employees posted photos of their screens on Instagram, and people found contact details and financial data sitting in the login page's code.
Perplexity open-sourced the sandbox that keeps its AI agents boxed in: in Perplexity's own testing, agents failed to escape it in 108 tries, though some slipped past its network rules before a fix.
Anthropic is reportedly offering Blackstone over 10% to help finance a $60 billion chip deal: borrowers with similar credit ratings usually pay about 7%, according to the Financial Times.
Learn more about AI from the experts building it
📸 Follow us on Instagram for fast, visual AI updates in 30 seconds.
📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.
🐦 Follow us on X for breaking AI news and real-time industry updates.
🧠 Learn how to build your next AI application with practical resources and expert guidance.
🎓 Start learning with free AI courses from GenAI Academy.
💰 Invest in the GTM infrastructure of the AI economy.
