ChatGPT Created a Fake Passport

A fellow named Boris Musielak managed to create a fake passport with OpenAI’s upgraded image model, generated by ChatGPT-4o.
He wrote that it took him 5 minutes to create a copy of his passport that most automated KYC systems would probably accept without hesitation. Don’t try this, because after the publication, OpenAI blocked this capability.
Today, it is much easier to fake anything, including IDs and documents, which raises many red flags, because there are financial websites that rely on photographing an ID card or passport, and with GenAI, it is also possible to fake video at a very high level. The only way to prevent counterfeiting is to use verified digital identity, such as eID wallets established by the European Union.

Another interesting case – a guy named Gavin Gaal created a restaurant meal receipt with it, including the business name, address, date, GST number (a unique identifier given to registered businesses) and detailed charges of what was ordered. He also asked for it to look a little crumpled, to look real. This is another opening for forging invoices that can be submitted to the tax authorities.

Apple’s Mulberry Project

Apple wants to make a significant change in the field of health. Its new project called Mulberry will include an AI-based health coach that will try to replace a service provided today by a doctor. Apple plans to bring in external experts, including doctors, to create video tutorials to accompany the service, in the areas of sleep, nutrition, mental health, physical therapy, and heart health. This new service will be launched in the US only due to regulatory changes in the US that restrict digital health services in most countries.

A Future Without Needles

I always asked myself why needles are still used in the field of health? And here, finally, we are seeing developments in this area. Three innovative technologies that can make needles obsolete:

1. Smart skin patches, equipped with tiny micro-needles, can release drugs directly into the bloodstream without pain. This technology is already being tested for administering vaccines and chronic medications.
2. Tiny capsules containing nanobots can release drugs in a targeted manner in the body, after ingestion. This technology allows precise delivery of drugs to the intestines or specific organs, while reducing side effects.
3. Smart inhalers will allow drug delivery through the lungs, not only for respiratory diseases but also for systemic conditions such as diabetes. The technology uses tiny particles that are absorbed directly into the bloodstream through the lungs.

These technologies are expected to change the way we consume drugs, and make medical treatment less invasive and more personalized.

Microsoft Copilot Upgrades

In honor of its 50th birthday, Microsoft upgraded Copilot. It will function like an AI agent and will be able to complete tasks such as browsing most websites, ordering tickets, it will remember your personal preferences such as favorite foods, music or movies, it will be able to analyze videos and photos and answer questions about them, and also track online transactions: for example, you can ask it to search for products on sale and it will update you when the price drops.

Doctor Robot

A Chinese surgeon successfully performed lung surgery on a patient 5,000 km away. The doctor removed a tumor from the lungs by operating a robot from a completely different city.

What is Agent2Agent?

Agent2Agent (A2A) is an open protocol that allows virtual AI agents to communicate securely, exchange information and coordinate actions across different platforms and applications.
It’s like a new language that allows different AI agents to talk to each other, even if they were created by different companies or in different frameworks. It’s like creating a super-efficient team of digital assistants that can work together seamlessly to perform complex tasks.
A2A was developed by Google in collaboration with more than 50 technology partners, including large companies like Atlassian, PayPal and Salesforce. The goal is to create a global standard that will allow AI agents to work together, regardless of platform or vendor.
How does it work in practice?
Imagine you need to find a software engineer. With A2A, you can ask the agent to find suitable candidates according to a list of requirements and skills. The agent will talk to other agents that specialize in recruiting, and they will find you potential candidates. Then, you can ask another agent to coordinate interviews. After the interview process is complete, another agent can perform background checks. All this happens automatically and quickly, thanks to the ability of the agents to communicate and work together. A2A allows them to work together seamlessly, like a team of people who collaborate. A2A is still in development, but already now it is possible to see its potential to change the way we work and live. In the future, AI agents will be able to collaborate to solve complex problems and improve our lives in many areas.

Innovations in ChatGPT

OpenAI is giving away ChatGPT Plus for free to students in the US and Canada, for the exam period, until May.

It has also created a free AI learning platform called OpenAI Academy.

From now on, ChatGPT can remember all our previous chats. Over time, it will better understand our personal preferences, and every interaction with it will be much more relevant to us. Sam Altman said that this is the beginning of a new era of building a unique relationship with a chatbot throughout our lives. Of course, this option can easily be turned off in the settings.

OpenAI has released several new models:

Model GPT-4.1 with a context window of one million tokens, which is capable of processing a large number of files (approximately 3,500 pages). This is a faster model, with a higher level of accuracy in analyzing information and answering complex questions. This model is currently available to developers through the API.

Model o3 is the smartest model that understands complex relationships, thinks and operates external tools. It excels especially in visual tasks such as analyzing images, charts and graphics. When it comes to complex tasks, it makes fewer significant errors (20% less) than model o1, especially in areas such as programming, business/consulting and creating creative ideas. It is available through the API.

Model o4-mini is a smaller model suitable for fast and cost-effective inference with the best performance in AIME tests. It outperforms the previous model o3-mini, which was replaced, in non-STEM tasks as well as in areas such as data science and supports processing a large amount of data.

Google Doesn’t Stop

Google is offering its advanced model called Gemini 2.5 Pro for free. It is a thinking model with improved performance and more accurate answers. It has the ability to identify and correct errors intuitively and automatically and its context window is 2 million tokens. It outperforms competitors in benchmarks such as AIME 2025 where the model’s capabilities are tested in mathematical tasks and GPQA Diamond designed for tasks in various fields of science.

It also launched a new development environment called Firebase Studio that allows you to build applications using text only. The tool uses Gemini 2.5 that writes code, tests it, fixes bugs and designs a user interface.

The Gemini Live Camera & Screen Sharing feature is available on Pixel 9 and Samsung S25 devices. This feature allows you to share the screen and use Gemini for all sorts of tasks that it can see through the camera or through screen sharing.

A new Google model called DolphinGemma allows you to talk to dolphins. This model was trained on dozens of underwater recordings, and it can identify patterns in dolphin sounds, and use them to create two-way communication.

I recommend watching a Google Cloud video that presents innovations in the world of AI, including the integration of AI agents that are already being integrated into organizations.

Recruiters and Lawyers – Your Profession is in Danger

Victor Lazarte, a partner at the venture capital fund Benchmark, claims that artificial intelligence not only helps with work but replaces human roles. According to him, fields such as law and recruiting are at high risk of automation. The claim is that artificial intelligence models will become more efficient than humans and perform tasks such as screening candidates and coordinating interviews. There are places where it also interviews the candidates. In the legal field, artificial intelligence may streamline routine tasks such as processing and analyzing legal documents (for lawyers) and integrating a whole world of information sources such as judgments, regulations, etc., to generate insights quickly and make better decisions.
This is the time to start thinking about tasks that artificial intelligence will not be able to perform in the future.

Quick Updates:

Motorola will integrate Perplexity’s voice assistant into the new Razr device, and it will work alongside Google Gemini and allow tasks such as booking flights without the need for dedicated apps.

Love Perplexity? From now on you can text and talk to it on WhatsApp: https://wa.me/18334363285

Two Chinese companies have received regulatory approval to operate autonomous flying taxis for short-distance passenger transport. These taxis reach speeds of up to 81 km/h.

How will we get around in the future? Check out a futuristic transportation concept with a cyborg horse named CORLEO

Meta introduced Llama 4 Scout and Llama 4 Maverick, new multi-modal artificial intelligence models capable of processing text, video, images and audio. These models are available as open source.

Midjourney has added an interesting innovation – “draft mode” that allows you to give voice instructions for creating images. It also understands voice instructions well for making changes to the images it created.

Samsung announced a new feature called Gemini Live on Galaxy S25 devices that allows users to get real-time help using the device’s camera. The feature provides recommendations and guidance based on visual analysis of the environment.

Claude offers a monthly subscription starting at $100 for anyone who wants to receive a significantly higher usage quota. Claude has added integration with Google’s Worlspace, including the ability to search documents, emails and calendar. These features are only available to paid subscribers.

Canva has released new AI-based capabilities – creating content and reports, new designs, building apps and more.

Netflix is ​​exploring a new search option powered by OpenAI that will allow users to find content according to their mood. The feature is currently available to iOS users in Australia and New Zealand, and in the coming weeks also in the United States.

YouTube is launching “Music Assistant”, an artificial intelligence tool that creates royalty-free instrumental tracks based on text prompts. The feature is gradually appearing as part of the Creator Music beta.

Researchers from MIT have developed a new method that allows training artificial intelligence models while maintaining their accuracy and preventing the disclosure of sensitive information. The method protects the privacy of the data and prevents attacks aimed at revealing confidential information from within the model.

Researchers at Bar-Ilan University have developed an artificial intelligence model that is able to predict with high accuracy forest fires caused by lightning. The model combines satellite data, weather conditions and other environmental factors, and offers an innovative tool for dealing with the effects of climate change.

To ensure a safer environment for teenagers, Meta is expanding the use of artificial intelligence to identify the age of users on Instagram. The system will examine signs such as birthday greetings and content preferences to identify users who are minors, and will automatically change their account settings.

Adobe announced that it will integrate artificial intelligence-based image creation models from OpenAI and Google into its Firefly app, which will also be available on mobile devices. The move is designed to expand creative options for users and allow content integration in apps like Photoshop in a simple way.

Make, an automation tool, has launched smart AI agents that understand the need and automatically build automation processes with the appropriate tools, which significantly streamlines the building of automation processes.