EU Enforces AI Act, Bans High-Risk Systems, and Imposes Strict Compliance Rules

The European Union has confirmed how regulators will now ban using AI systems that pose high levels of risk or danger as per the latest AI Act.

February 2 was outlined to be the latest deadline on this front, featuring a comprehensive framework on what’s acceptable and what’s not. This law came into sight on August 1 and now its list of compliance deadlines is taking center stage.

The details outlined here are curated to include several different use cases where the technology could interact with different people. For instance, it might be a consumer application or a physical environment.

Through this latest bloc technique, four risk levels were identified. This includes minimal risk which will not feature any kind of regulatory oversight. Then limited risk is designed to entail customer service chatbots featuring light-weight regulatory oversight. In the end, high risk is for those featuring AI for serious life-changing matters including health. These will in turn have to deal with strict regulatory oversight. There is similarly a category for unacceptable risk that focuses more on compliance requirements that would be eliminated as a whole.

There were certain examples provided about behaviors deemed unacceptable. This includes social scoring, manipulating decisions, exploiting user vulnerabilities, and predicting crimes depending on physical appearance.

AI using biometrics to interfere with an individual’s characteristics such as gender and collecting real-time data in public places for the sake of law enforcement is similarly not acceptable. Any AI inferring emotions in a public place setting and those scraping pictures online through security devices such as cameras will similarly be eliminated.

Any company found guilty of being involved in the above AI applications will have to deal with fines, no matter where they might be located. This means it could give rise to bigger fines that go up to $36M. As of now, no date for when the fines would implemented was shared.

However, all companies need to be compliant by February 2 and the next major compliance deadline would be August for this year. This way, users will be aware of competent individuals and how enforcement provisions will come into play.

Some deem this latest deadline of February 2 as a formality. More than 100 different companies laid down a voluntary pledge to begin applying AI Act principles into this application. Signatories such as Amazon, OpenAI, and Google were said to be committed to highlighting any risks inside the systems that they feel to the greatest degree. However, it was interesting to note that several others such as Mistral, Apple, and Meta skipped being involved in this pact and decided not to sign.

So what does this mean? The companies that didn’t sign vow not to engage in such practices anyway. They are already mindful of the actions and which cases are prohibited as per experts.

This latest AI Act will not be working independently in the region. It will interact with the new AI Act but the question on many people’s minds is how the different laws will fit together and work as one. After all, they all have their own set of defined clauses coming into existence in one place.

Image: DIW-Aigen

Read next: OpenAI Announces New AI Agent Designed to Assist People Carryout In-Depth and Complex Research
by Dr. Hura Anwar via Digital Information World

OpenAI Announces New AI Agent Designed to Assist People Carryout In-Depth and Complex Research

Tech giant OpenAI is releasing a new AI agent that was designed to assist users carry out in-depth and complex research through ChatGPT.

The latest offering dubbed Deep Research is said to be for those seeking extensive knowledge in the domain of science, finance, engineering, and policies. It’s also said to be useful for anyone who needs to make a big investment but needs more in-depth analysis to get that done.

In its latest blog post, the company shared how the latest capability was curated for those doing intensive knowledge such as a vehicle or furniture. So in other words, it’s created for situations where you require an instant summary or reply but instead require more detailed data from several different sources.

OpenAI says it’s giving the latest model to those with ChatGPT Pro today, limited to a few queries every month. After this, it hopes to share the model with users having Team and Plus, and also Enterprise. The tech giant also shared how the launch is geo-targeted. The company had zero release timeline to share for those located in the UK region, including Switzerland, and the EU.

To better use the tool, you just need to select the term Deep Research through the drop-down menu and generate a question. You then have the option to attach files or any spreadsheet. It might take 5 to 30 minutes to provide a reply, and users will be presented with an alert when their search is done.

For now, the tool’s outputs are based solely on text. However, the company does intend to include embedded pictures, analytic outputs, and more data visualizations. Meanwhile, the roadmap also features the option to link more intricate data sources such as those linked to subscriptions and internal sources.

The biggest query right now is levels for precision. AI is more than perfect and could give rise to more hallucinations and different errors that might be dangerous in scenarios related to deep research. This is why the tool will document every output when carrying out deep research. It will entail summaries of the thinking process and outlined citations clearly. This helps in the verification of data and also in the referencing process.

We’ve already seen the ChatGPT on web searches make frequent errors and give misinformation. Furthermore, tests showed how it gave rise to less reliable results in comparison to Google Search for any questions.

The company tried to defend the latest model’s accuracy with the use of o3 reasoning for different AI models. It was also trained through reinforced learning techniques on tasks done in real-time.

Reinforcement learning assists the model through trial and error techniques. As it gets near the goal of producing an answer, it gets virtual rewards to make it more competent for future tasks.

Read next: New Survey Shows Goals and Challenges Small Businesses Have and What they Think of AI
by Dr. Hura Anwar via Digital Information World

New Survey Shows Goals and Challenges Small Businesses Have and What they Think of AI

According to recent research from Taradel, small businesses are currently facing challenges such as limited budgets and lead generation difficulties. The report conducted a survey among 226 small businesses and found out what top challenges they are facing, what are their goals and what do they think about AI usage. 27% of the small businesses surveyed said that budgeting is their main challenge right now while 30% said that driving lead generation is the top marketing challenge they are facing in their business.

On the other hand, talking about their business goals, 55.6% of the small businesses said that prioritizing increasing sales is their top goal right now, with 19.2% said their main goal is to generate leads. 12% said that they are aiming to build their brand awareness right now. Reducing costs, retaining existing customers and launching or growing ecommerce business are also some of the main goals defined by small businesses which were surveyed. Talking about generating leads, 24% prefer phone calls while 22% small businesses prefer in-business consultations.

63% of the small businesses said that social media platforms are a great way to maintain their online presence. Facebook and Instagram are top social media platforms for small businesses while email remains the number one offline platform for small business marketing. Small businesses were also asked about their thoughts on AI and 43% said that they do not use AI tools at all. 34% said they use AI tools a little while 18% said they use AI tools extensively. Most small businesses surveyed said that they are using AI for content creation, email marketing and social media management. Only 9% of the small businesses said that their marketing tactics are highly effective while 21% said their marketing tactics are not effective at all.

Read next: Global AI Safety Report Warns of Cyber Threats, Manipulation, and Weaponization Risks
by Arooj Ahmed via Digital Information World

Smartphones Under $600 See 20% Drop as Demand Shifts to Higher-End Devices

According to Counterpoint's US Smartphone Tracker, there was a 3% decline in US smartphone sales in Q4 2024 which was mostly due to low sales in the sub-$600 smartphones and devices priced below $300. However, a big reason for decline was also because of a 4% YoY increase in sales of smartphones above the $600 price segment because there was an increased demand for premium devices during holidays. Maurice Klaehne, Senior Research Analyst, said the US smartphone market was dominated by premium smartphones in Q4 2024 with iPhone 16 series seeing a 1% YoY increase in its sales as compared to the series like iPhone 15. Consumers are opting to upgrade to premium and expensive smartphones as well.

Google also saw a good performance in Q4 2024, mostly due to the Pixel 9 lineup which saw double digit growth. Almost a year after the launch of GenAI smartphones in the US, demand remains muted. However, it is expected to grow with the release of second-generation AI-powered devices like the S25. The sub-$600 smartphone market saw a 20% YoY decline because of the low end market and weaker demand, especially for devices under $300. Consumers in this segment are more price-sensitive and have not benefited from the broader macroeconomic recovery in terms of job and wage growth..

• Also read: Average Smartphone Price Hits $356 as Premium Demand Grows in 2024

Research Director of Counterpoint, Jeff Fieldhack said that carriers strongly impacted Q4 2024 by offering promotions like bundles with tablets, watches or broadband services. This helped them in increased upgrade rates and equipment revenues as consumers opted for high priced devices. But channels like prepaid and Lifeline saw decline but carriers are still optimistic about growth in 2025, especially Verizon with unique promotion strategies. With replacement cycles peaking in 2024, carriers expect higher smartphone sales in 2025 as more consumers move past their contract periods and look to upgrade.

Read next:

• Study Reveals Students Turn to AI Chatbots for Emotional Support Amid Rising Loneliness

• Are GPT-2’s Impressive Download Numbers a Sign That Older Models Still Have Market Power?
by Arooj Ahmed via Digital Information World

Study Reveals Students Turn to AI Chatbots for Emotional Support Amid Rising Loneliness

According to a new study published in International Journal of Human Computer Studies, some students use AI chatbots for emotional support and social conversations. The students who do this show high levels of loneliness and lower social support so they turn towards AI chatbots to fill the void. There are many social chatbots like Character AI and Replika that people use for conversation in an engaging, stimulating and empathetic manner. These social chatbots have shifted how humans communicate and as there will be more advancements in them, our socio-cultural environment is going to change as well. This change will be mostly seen in young teens.

The study aimed to provide answers to some questions like why teens are engaging in social conversations with chatbots, how many of them do so and if it is all linked to less social support and more loneliness. For the study, the researchers conducted a survey of 1,599 students from 15 Danish schools and different qualitative and quantitative methods were used for it. The main focus of the study was to find whether students are engaging in friend-like conversations with chatbots, emotionally and casually.

The results of the survey showed that 14.6% students reported having friend-like conversations with chatbots but most of the interactions were utility based and not social. Only 2.4% of students reported having social or emotional conversations with AI chatbots. This shows that this phenomenon is not widespread as many people think so but we cannot say whether it is going to increase in the future or not.

The students who were having emotional conversations with chatbots showed signs of loneliness and they were less supported by their peers as well. Whenever these students are in a lonely, angry or bad mood, they turn to chatbots to share their feelings and use them as coping tools. But as students are turning to chatbots in times of loneliness, this raises concerns about their emotional well being. It is not clear whether chatbots are a cause of loneliness or students or if lonely students are turning towards chatbots during times of loneliness.

There are some limitations in the research too, like it was based on self-participation bias and no clear definition of friend-like conversation was given. There should be more methods to understand the phenomenon of students using social chatbots and more research should be done to understand why they use them.

Image: DIW-Aigen

Read next: New Study Shows Bots Can Increase Online Engagement but This Can Decrease Human-to-Human Interaction
by Arooj Ahmed via Digital Information World

Sam Altman Admits OpenAI’s Closed Strategy Is Flawed as DeepSeek Disrupts AI Market

Sam Altman made a shocking admission yesterday, referring to his OpenAI company as being on the wrong side of history. This is regarding the subject of open-source AI. The news is awfully interesting as it proves a major shift in terms of strategy.

Competition from China continues to heat up and so many open models are being launched with incredible capabilities and at a fraction of the cost. The very candid session opened many people’s eyes in terms of DeepSeek rattling the global markets of today with the new R1 model. It is believed to be offering great performance and promising results, allowing users to shift focus from the ChatGPT makers and move on to better alternatives.

Altman did share how he personally feels the company might be on the wrong side of history and therefore needs to think of a different open-source strategy. Remember, not all share the same views as Altman, and right now, this is definitely his company’s top priority at the moment.

The statement is a representation of a major leap from the proprietary approach taken up in recent times. This drew serious backlash from many AI experts and ex-allies including SpaceX CEO Elon Musk. The latter is suing the firm for trying to betray the real open-source mission.

Altman’s comments on this matter come at a time when the market is experiencing serious turmoil thanks to DeepSeek. Using just $5B in training, it saw many tech giant stocks going down including Nvidia. To be more exact, a whopping $600 billion worth of market value was instantly wiped out which is a major drop for any firm working in the US.

As per the organization, it would seek to produce better models but does feel the lead would be far less than what it had in the past thanks to the breakthrough impact of DeepSeek. Right now, Altman feels the company’s closed strategy is more flawed than anything else.

The success of DeepSeek gives rise to a major shift in AI dynamics. The fact that it could give rise to startling results with just 2000 Nvidia H800 GPUs is remarkable. On average, 10,000 chips and more are used by most leading AI labs. Hence, the results do imply that innovation in algorithms and optimizing architecture could mean more than just raw computing power.

Furthermore, such a revelation threatens not only the firm’s tech strategy but the complete business model designed for exclusive access to huge computational resources.

The debate has to do with innovation versus the security of open-source models. The matter keeps getting intensified with time with many firms storing user data on different servers across China. This means government officials have full access. We’ve already seen lawmakers in the US speak about restrictions on this move and NASA is the latest to block the app, linking major privacy violations.

Image: DIW-Aigen

Read next: OpenAI’s o3-Mini Debuts with Enhanced Reasoning, Challenging DeepSeek R1
by Dr. Hura Anwar via Digital Information World

OpenAI’s o3-Mini Debuts with Enhanced Reasoning, Challenging DeepSeek R1

It looks like OpenAI is battling the rising success of DeepSeek’s latest R1 model by debuting a new o3-mini model. However, the question is if it will be good enough to bar the Chinese startup giant’s success.

The launch comes a few days after we heard growing rumors about the rise in anticipation among AI users through social media. This is the company’s second model that comes under the reasoning category of models. In other words, the company says it can take more time to evaluate situations and reflect on its chain of thought. This is before it jumps ahead of gives answers to prompts put in by the user.

The final result is a model that performs at the same level as those having a PhD degree or those giving replies to hard queries such as math, engineering, science, and more. The latest model is now up for grabs through ChatGPT and also entails a free-of-cost tier. You can find it through the company’s API.

Thankfully, the model is cheaper in cost, quicker in terms of speed, and also able to perform better than models of a similar kind seen in the past. This includes its own sibling o1-mini. While it can be compared to the DeepSeek R1, many don’t know that this is a planned release for the company. It was shared well before the launch of DeepSeek’s R1 so you can’t actually deem it as a strategy by the company. To be exact, it was declared in December 2024 and that’s when Sam Altman also shared through his social media profile that it was under discussion and would launch together on ChatGPT and the OpenAI API together.

This mini model won’t be up for grabs in terms of an open-source model. This means codes can’t get downloaded for usage offline nor could they undergo customization to a similar extent. However, it could restrict appeal when compared to the DeepSeek-R1 for a few applications.

The company failed to roll out any more details about the bigger o3 model shared in December with the o3-mini. During that moment, it shared a delay of several weeks before testing with third parties.

In terms of performance and its features, it’s quite like o1. The OpenAI o3-mini is designed to give reasoning in coding, science, and math. The performance is similar to OpenAI o1 in terms of reasoning but it does offer great advantages.

There’s a 24% quicker response time when you compare the o1-mini speed time for replies is down to just 10.3 seconds. You can expect better accuracy and external testers putting the mini model replies 56% of the time.

Meanwhile, it also provides 39% fewer errors on complex queries. There’s similarly greater performance in coding and performing STEM tasks, not to mention good reasoning effort. Users can benefit from reasoning done on a low, medium, and high level. It similarly gives users the chance to balance both speed with reliability.

The model even has great benchmarks including outpacing o1 in plenty of cases as per the o3-mini system card shared by the company online. The context window is 200k and has a maximum of 100k in every output. It’s quite like the complete o1 model that edges out DeepSeek’s content window for R1. However, it’s yet to outperform Google Gemini 2.0 Flash Thinking model. The latter can produce a higher performance that goes up to 1M tokens.

Reasoning capabilities might be there but the model does fall behind in terms of vision capabilities. This is why any user wishing to upload pictures could be better off with the o1 for now.

Read next: Global AI Safety Report Warns of Cyber Threats, Manipulation, and Weaponization Risks
by Dr. Hura Anwar via Digital Information World