DeepSeek's First Problems | Weekly Edition

PLUS HOT AI Tools & Tutorials

Jan 31, 2025

Hello and welcome to our weekly roundup!

It's been over a week since DeepSeek's release, and the talk about it is still going strong. However, in recent days, widespread excitement has been replaced by skepticism amid a string of problems.

Today, we deal with the first challenges faced by the Chinese startup and other AI events of the past seven days. Let’s get started!

If you haven't figured out DeepSeek R1 yet and want to know more about it, we have a suitable post for you:
DeepSeek R1: Everything You Need to Know
Stepan Ikaev
·
Jan 28
Read full story

This Creators’ AI Edition:

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁

Your advertisement could be featured here!

Sponsor a spot in our newsletter to connect with subscribers interested in AI, technology, and startups. For inquiries, click the button below.

Featured Material 🎟️

DeepSeek Faces First Challenges

So, after a few days of hype, DeepSeek suffered some losses, both reputational and actual. Let's briefly analyze all the events related to the Chinese company.

OpenAI vs DeepSeek

The first, but not the most significant, event was the conflict with OpenAI. The company has accused DeepSeek of using its proprietary technology without permission. Specifically, OpenAI suspects that DeepSeek employed a technique called "distillation" to enhance its AI model by learning from larger, more advanced models, possibly breaching OpenAI's terms of service.

Microsoft also found that individuals linked to DeepSeek might have harvested significant amounts of data from OpenAI's API, prompting an ongoing investigation. Whether OpenAI or Microsoft plans to enter a legal fight with the Chinese startup is unknown. However, this has already affected DeepSeek’s reputation.

Keep your mailbox updated with key knowledge & news from the AI industry

Leak of User Data Into the Public Domain

The second event proved to be a bigger blow for DeepSeek.

Wiz researchers found a publicly available ClickHouse database owned by DeepSeek that allowed complete control over database operations, including access to internal data. The company said it took “minutes” for experts to find all the information.

The exposure included more than a million rows containing chat history, secret keys, backend details, and other user data—all without any authentication or defense mechanism for the outside world. DeepSeek secured the database after Wiz notified the startup of the problem.

Interestingly, Wiz also told Wired that DeepSeek's systems are designed similarly to those used by OpenAI “down to details like the format of the API keys.”

DeepSeek’s Ban in Italy

And the third one.

Italy's Data Protection Authority, Garante, has blocked DeepSeek due to insufficient information regarding its data usage. The app disappeared from Italian Apple and Google app stores after Garante sought details on the types of personal data collected, its sources, purposes, legal basis, and whether it is stored in China.

According to the office, the blocking decision, which takes immediate effect, aims to protect Italian users' data following incomplete responses from the Chinese suppliers of DeepSeek’s chatbot service. An investigation has also been launched.

Something tells me if DeepSeek keeps the same policy, it could be blocked not only in Italy but also in other countries.

News Of The Week 🌍

But There's Janus-Pro AI Image Generator

One positive news about the same startup.

Earlier this week, DeepSeek unveiled its latest open-source text-to-image model, Janus-Pro-7B. It enhances training processes, data quality, and model size, resulting in more stable and detailed images. The developer says that by incorporating 72M high-quality synthetic images alongside real-world data, Janus-Pro-7B achieves nice image outputs. Despite this, the first tests don’t show the most impressive results.

Below is a comparison of JanusPro with other popular models.

WSJ: OpenAI Gets Ready to Raise $40B

OpenAI is reportedly in advanced talks to raise $40B in a funding round led by SoftBank, which could elevate its valuation to $300B. This would position OpenAI as one of the most valuable private companies globally, second only to SpaceX. $25B will go towards OpenAI's unprofitable operating expenses, and the remaining $15 will go towards the ambitious “Stargate” project.

If the deal is confirmed, Softbank will become OpenAI's leading investor, surpassing Microsoft.

Jack Dorsey’s AI Startup Introduces Open Source AI Agent

The wave of AI agents doesn't even think about slowing down.

Jack Dorsey's (founder and ex-CEO of Twitter) startup launched Codename Goose, an open-source AI agent designed to automate tasks, particularly in software development. Goose is a platform that integrates with various LLMs, including GPT, Claude, and DeepSeek-R1. It is built to enhance productivity by automating coding tasks and (according to Block) can also be adapted for non-engineering purposes in the future.

Goose is available under the Apache 2.0 license, allowing both academic and commercial use. You can run it locally.

Alibaba's Latest AI Model Promises to Surpass DeepSeek

DeepSeek isn't the only Chinese company that wants to be a leader in AI. Alibaba recently unveiled its latest AI model, Qwen 2.5-Max, claiming it outperforms leading models like OpenAI's GPT-4, Meta's Llama-3.1-405B, and DeepSeek-V3 in various benchmarks. The model is trained on over 20 trillion tokens and employs a Mixture of Experts architecture for specialized task handling.

Unlike other Alibaba models that have been released as open source, Qwen 2.5-Max is closed source. So, at least in this regard, this AI loses out to DeepSeek.

Sharing is caring! Refer someone who recently started a learning journey in AI. Make them more productive and earn rewards!

Refer a friend

Google Gemini 2.0 Flash Now Available to All Users

Google is trying to keep up with its competitors in the battle for a wide audience. This week, the company announced that the Gemini 2.0 Flash model is now available to all registered users. You can try it out on both the website and the Gemini mobile app. As Google says, the 2.0 Flash is a balanced AI between the performance of Pro and the compact of Nano.

If you install Gemini as an assistant on your smartphone, Gemini 2.0 Flash will also be responsible for its operation.

Microsoft Makes OpenAI’s o1 Free for Copilot Users

Another big company has made AI more accessible. Microsoft has announced that its AI feature, "Think Deeper," is now available for free to all Copilot users. This feature, powered by OpenAI's o1 reasoning model, enables Copilot to tackle complex questions by analyzing them from multiple perspectives and providing step-by-step solutions. Initially launched in October for Copilot Pro subscribers, this feature could benefit tasks such as comparing options, coding, or planning.

You won't need a $20/mo subscription to Copilot, either.

Useful Tools ⚒️

Kiva - AI-powered SEO agent for agencies, SMEs & startups

Rocktangle – AI tool to automate market research (saves 10 hours/week)

Remy – Ask any question and get answers from the world’s videos

GPTSeek – Export chats from ChatGPT to DeepSeek Chat in 1-click

Stella by FastTracker AI - Automate meeting invites, emails, and note-taking

Stella is a new AI platform for task automation. Its main feature (and the difference from other assistants) is that this tool does not work within a separate app but is integrated into WhatsApp. So you don't need to download anything or log in. Stella generates draft emails, helps you keep a working calendar, tracks reminders, and translates voice notes into text.