Skip to main content

Google Uses Geothermal Technology to Vaporize Rock for Power

Google is now using geothermal technology to vaporize rock deep underground. This project aims to create a steady supply of clean power for the company. By drilling deep and using high heat, they can tap into energy that never runs out. How does Google geothermal technology work? Traditional green energy like wind and solar only works when the weather is right. Geothermal technology is different because it uses the heat that is already inside the Earth. Google is working with partners to drill deeper than ever before to reach extremely hot temperatures. This process involves turning water into steam by touching hot rocks or even vaporizing the rock itself to release energy. The company wants to move away from fossil fuels that hurt the planet. Using the earth as a giant battery is a smart way to get power without making smoke. This new method allows them to build power plants in places where it was once impossible. They are pushing the limits of what machines can do in high heat en...

EU AI Act Rules and Deadlines That Tech Companies Face

EU AI Act Rules and Deadlines That Tech Companies Face

The European Union has set hard deadlines for the EU AI Act, and tech firms must follow them or pay huge fines. European officials passed this broad law to set guardrails on software that recognizes faces, sorts job applications, and builds synthetic text. If you build software or run an online tool, these new requirements will touch how you work.

The short answer is simple. The EU AI Act sorts artificial intelligence tools into risk groups, completely bans a few of them, and demands clear testing and reporting for the rest. Companies that break the rules can face fines of up to 35 million euros or seven percent of their global annual sales, whichever is higher.

Most small teams think rules like this only apply to giants like Google or Microsoft. That is not how the text is written. Anyone offering software inside the European market has to follow along, even if their servers sit in Texas or Tokyo. Let us break down what is happening, what gets banned first, and what you need to track.

What Is the EU AI Act and Who Must Follow It?

The European Union created the AI Act to give buyers and citizens protection from biased or dangerous software systems. Lawmakers worked on the draft for three years before the European Parliament voted to approve it in the spring of 2024. You can read the official timeline on the European Parliament website.

The law covers anyone who sells, hosts, or imports computer programs that use machine learning or algorithmic decision systems in EU countries. If your app has active European users, the location of your business headquarters does not matter. The law looks at where the end user lives. A developer working alone in Toronto faces the same rules as an enterprise team in Berlin if their tool touches European consumers.

Not every computer program counts as artificial intelligence under this framework. Simple scripts that follow fixed if-then rules do not fall under the strict sections. The act targets programs that infer ways to achieve goals, adapt their outputs, and process inputs like text, audio, and visual clips without explicit manual steps.

If you keep up with broader tech shifts on our PromptNova homepage, you know that keeping track of legal updates is just as important as watching new software updates. The rules create four clear risk groups: unacceptable risk, high risk, specific transparency risk, and minimal risk.

Which AI Systems Are Completely Banned?

The first deadline bans tools that the European Union labels an unacceptable risk. These bans take effect six months after the law enters the official journal. Officials decided that some uses cross basic human rights lines, so no company can offer them within member states under any circumstances.

Social scoring systems by governments or private businesses are totally illegal. If a bank or local council wants to track your daily shopping and online habits to give you a public trust score, the law blocks it. China uses social scoring tools, but the European Union wants to stop that model from ever taking root in its cities.

Real-time remote biometric identification in public spaces by police also faces heavy bans. There are very narrow exceptions, such as searching for a kidnap victim or preventing an immediate terror threat. Even then, police officers must get permission from a judge before turning on public facial scanning cameras.

Cognitive behavioral manipulation that targets vulnerable groups is another banned category. An example would be an audio toy that uses voice prompts to urge a young child to perform dangerous acts. Scraping facial images from security footage or the public web to build facial databases, like Clearview AI did, is also forbidden under the text.

Emotion recognition software inside schools and workplaces is banned too. Your employer cannot run webcams to scan your face and guess whether you feel happy, tired, or angry during your shift. The lawmakers decided that these systems are both unreliable and intrusive.

How Do High Risk AI Rules Work?

The largest part of the law focuses on high risk systems. These are tools that do not face total bans, but can still harm a person's health, safety, or basic civil rights if they fail. Companies building high risk software have to complete heavy compliance steps before launching.

What counts as high risk? Here are the main areas named in the legal text:

  • Hiring tools that filter resumes or rank job applicants during video interviews.
  • Credit scoring tools that decide if a family can get a home loan.
  • Educational software that scores standardized tests or decides school admissions.
  • Healthcare software used to help doctors triage emergency patients.
  • Critical infrastructure controls for water, power, and transport networks.
  • Law enforcement tools that assess the risk of a person committing a future crime.

If your tool falls into one of these buckets, you cannot just push code to a server. You must set up a full risk management system that runs throughout the tool's lifecycle. You have to train your models on high-quality data to reduce bias, keep detailed technical logs, and make sure human workers can stop or override the machine at any moment.

Companies must also register their high risk systems in a public European database. This lets watchdogs and consumers see who built the tool, what training data it used, and how the team tests for errors. If you follow modern generative video tools like the OpenAI Sora AI Video Model Features and Release Context, you know that tracking data inputs and system limits is already becoming standard practice for responsible teams.

What Are the Rules for ChatGPT and General Purpose AI?

When the European Union first started writing this law in 2021, large language models were not widespread. The sudden boom of tools from OpenAI, Google, and Anthropic forced regulators to rewrite whole sections to address general purpose AI models. These are base models that can do hundreds of different jobs rather than one single task.

The rules split general purpose models into two tiers. The first tier covers any standard base model. The creators must provide technical documentation, follow European copyright laws when gathering web data, and publish a detailed summary of the content used to train the system. This gives artists and publishers a way to see if their books or articles were copied into the model.

The second tier covers models that carry systemic risk. The law uses a specific computing metric here: any model trained using more than 10 to the power of 25 floating point operations, or FLOPs, hits this tier. The biggest models from major tech firms pass this line today. You can read technical breakdowns of compute thresholds on research pages like the Wikipedia entry on the Artificial Intelligence Act.

Makers of these massive models must run model evaluations, perform red-teaming tests to find security holes, track and report serious incidents to European regulators, and ensure strong cybersecurity protections. The European AI Office, a new regulatory body set up in Brussels, directly oversees these model makers.

Open source models receive some exemptions, but only if they stay below the systemic risk threshold and do not fall into high risk categories. If an open source model releases all its weights and code publicly, it skips some paperwork, but it still has to respect copyright laws.

Watermarks and Deepfakes: What Must Be Labeled?

Have you ever seen a photo or video online and wondered if it was real? The EU AI Act addresses this directly by requiring clear watermarks and labels on synthetic media. Any system that generates synthetic audio, photos, or video clips must mark those files in a machine-readable format.

If you publish a deepfake that shows real people doing or saying things they never did, you must disclose that the content was artificially created. The disclosure must be clear and visible to anyone viewing or listening to the material. An exception exists for obvious parody or artistic work, but creators must still display the label so users are not misled.

Text generators face disclosure rules too. If an online service uses an AI chatbot to talk to customers, it must inform people right away that they are speaking with a machine, unless that fact is obvious from the context. News articles and public interest texts generated by software without human editing must carry similar notices.

This is a big change for media outlets, marketing agencies, and social platforms. Watermarking standards are still being refined by groups like the Coalition for Content Provenance and Authenticity, but tech teams must get their technical tagging ready now to comply with the enforcement dates.

The Official Deadlines You Need to Track

The European Union did not turn on all rules in one day. Instead, they staggered the enforcement over three years so businesses have time to adapt. Here is the exact schedule you should mark on your calendar:

  • 6 Months: Bans on unacceptable risk systems take effect. Emotion tracking at work, social scoring, and untargeted web scraping for faces must stop.
  • 12 Months: Rules for general purpose AI models begin. Model creators must follow copyright rules and share model training summaries. Governance authorities take their seats.
  • 24 Months: Most high risk AI rules become fully enforceable. Systems used in schools, jobs, loans, and public services must meet all data and testing standards.
  • 36 Months: High risk systems that are built into products already covered by older EU safety laws, such as medical devices, cars, and airplanes, must comply.

Missing these dates carries severe costs. Small businesses and startups get slight discounts on penalties, but fines can still reach 15 million euros or three percent of global revenue for general violations. Supplying false or misleading information to regulators carries fines of up to 7.5 million euros or 1.5 percent of sales.

The European Commission provides guidance documents and sandbox environments where developers can test their tools safely before launching. These sandboxes help small teams check if their software meets standards without fearing immediate fines while testing.

How Tech Teams Should Prepare Right Now

You do not need to hire an expensive law firm tomorrow, but you do need an accurate inventory of the software tools your team uses. Start by listing every algorithm and third-party API in your product stack. Determine whether your tool handles sensitive decisions like hiring, identity verification, or user profiling.

Talk to your vendors. If your app relies on outside models, ask those providers for their compliance roadmaps. Under the law, if you take a base model from another company and modify it for a high risk purpose, you might inherit the legal duties of a provider. That means you become responsible for technical files, logging, and safety testing.

Set up clean records for your training data. Keep logs showing where you got your text, images, or code. Document how you cleaned the data and what steps you took to prevent bias against protected groups. Having clean paperwork ready will save your team months of frantic work when European regulators start asking questions.

Watch for updates on the official European Commission digital policies page. They publish ongoing codes of practice that explain how to meet the rules in day-to-day coding work.

What This Means for the Future of Technology

Some critics claim these strict rules will push tech innovators out of Europe. They point to delays in model releases across the continent as evidence that red tape slows down development. On the other side, consumer groups argue that clear guardrails build trust, making ordinary people more comfortable using automated tools in daily life.

We saw a similar story play out with GDPR privacy rules in 2018. Many companies complained at first, but the rules eventually became a global standard that shaped privacy laws in California, Brazil, and across Asia. The EU AI Act will likely have the same global reach. If you want to sell software anywhere in the developed world, building to European standards will soon be the easiest path.

Whether you run an independent app, manage an enterprise engineering team, or read tech news to stay informed, watching these rules matters. The era of unregulated automated systems is closing. Building clean, honest, and reliable software is no longer just good engineering practice; it is now the law.

Keep following our guides and updates on the PromptNova main hub to stay ahead of upcoming regulations, model releases, and technical shifts. The best time to check your code is before the deadlines arrive.

Comments

Popular posts from this blog

PromptNova2027: How to Build Your Ultimate Content Engine

Do you feel tired when you write blog posts? You are not alone in this. Many people want to share their thoughts online, but they run out of time. Creating good text takes hours. Sometimes, it takes days. That is why you need a system to help you work faster and better. We call this system a content engine. It is a simple way to make writing easy and fast. In this guide, we will look at how you can build your own content engine. We will use the PromptNova model to show you how. This model helps you write faster without losing your personal style. At PromptNova, we love finding ways to make technology work for you. You can visit AI Tools & Technology to see our latest guides and tips for online creators. If you want to build a system like this, you can explore this related resource to get started with automation. This will help you understand how simple tools can save you hours of work each week. What Is a Content Engine? A content engine is not just a single piece of so...

Become a Claymation Character Today

Have you ever watched those cute clay movies and wished you could be in them? Think about Wallace and Gromit or Chicken Run. The characters look so real and funny. Now you can join that world with the help of AI image tools. People everywhere are using AI to change their normal selfies into amazing clay figures. It's one of the most popular trends on social media right now. In this post, you will learn how to turn your own photo into a cute claymation masterpiece. You don't need any special art skills to do this. All you need is the right words and a good photo. We'll show you the exact steps to get this look in just a few minutes. What This Style Looks Like The claymation style has a very special look that people love instantly. It looks like someone spent hours shaping a piece of clay with their hands. When you use this style, your photo will get a soft, matte texture. You might even see tiny thumbprints or small lines in the clay, which makes it look very real. ...

AI Agents vs AI Chatbots: What Is the Difference?

Do you feel confused by all the new tech terms? You are not alone. Every day we hear about new tools. First, we had chatbots. Now, everyone is talking about agents. What is the actual difference between them? It is easy to think they are the same thing, but they are not. Let's look at how these tools work so you can choose the best one. To understand where technology is going, we must look at how these systems handle tasks. Chatbots talk to you, while agents do things for you. Many people want to find the best way to use these tools for work or daily tasks. If you want to find more options, you can check out useful AI resources to see how these new systems work in real life. This will help you see the difference clearly. What Is an AI Chatbot? An AI chatbot is a program designed to talk with human users. You type a prompt. The chatbot gives you a response. It uses language models to predict the best words to say next. This process is very fast. It feels like you are tex...