Meta in Panic Mode? DeepSeek’s AI Breakthrough Sends Shockwaves Through Silicon Valley

Meta has set up four internal teams to figure out how DeepSeek, a small Chinese AI startup, managed to roll out an AI assistant that’s already being called game-changing and impressive. DeepSeek’s latest chatbot model, R1, is said to be on par with top tier AI models like ChatGPT but at a fraction of the cost. The newest large language model on the block not only optimized compute usage but also open sourced the model which makes the competition in the AI space even more intense.

Industry insiders think Meta’s Llama models might have been the inspiration for DeepSeek. Given Llama is open sourced and so widely used, it’s possible some of the design elements were borrowed. But the performance and cost efficiency of the new Chinese model has Meta surprised.

According to insider reports, Meta’s AI infrastructure director, Mathew Oldham, has expressed concerns internally that DeepSeek’s model might even surpass the forthcoming iteration of Llama AI. This has put Meta in a race to close the gap before its own next-generation system arrives, which CEO Mark Zuckerberg previously hinted could launch in early 2025.

Inside Meta’s Response Strategy

Among the four specialized teams Meta has deployed, two are focused on deciphering how High-Flyer Capital Management - the hedge fund backing DeepSeek - managed to drastically cut training and operational costs for the model. The objective is to identify cost-reduction strategies that could be integrated into Meta’s AI projects.

A third team is examining the dataset DeepSeek used to train its model, aiming to understand whether unique data sources contributed to its efficiency. Meanwhile, the fourth group is assessing potential structural improvements for Llama based on DeepSeek’s architecture.

• Also read: Meta's Cross-Platform AI Memory Feature Rolls Out, But Privacy Fears May Hinder Adoption

Despite the competitive challenge, Zuckerberg has not publicly addressed DeepSeek’s rapid emergence. However, in a recent Facebook update, he reaffirmed that Meta’s upcoming Llama iteration would set a new industry benchmark upon release. He also disclosed plans to allocate $65 billion toward AI advancements in 2025, underscoring the company’s commitment to staying ahead in the generative AI race.

Meta’s Leadership Reacts

Yann LeCun, Meta’s chief AI scientist, addressed concerns on LinkedIn, maintaining a composed stance. He argued that DeepSeek’s advancements should not be viewed as a sign of China surpassing the U.S. in AI but rather as a testament to the power of open-source models outperforming proprietary alternatives.

He emphasized that DeepSeek built upon existing open research, demonstrating the collaborative strength of open-source innovation. In his view, the ability of researchers worldwide to iterate on shared knowledge benefits the broader AI ecosystem.

With the AI landscape evolving rapidly, Meta now finds itself in an urgent race - not just to understand DeepSeek’s breakthrough but to ensure its own future models remain competitive in a space where cost efficiency and open-source strategies are increasingly shaping industry leadership.

Image: DIW-Aigen

Read next: 

• Is YouTube’s Ad Moderation Failing to Protect Children from Inappropriate Content?

• Which Cryptocurrencies Are Americans Most Likely to Invest in for 2025?

• Mobile App Usage Surpasses 4.2 Trillion Hours in 2024, Revenue Growth Continues

by Asim BN via Digital Information World

Mobile App Usage Surpasses 4.2 Trillion Hours in 2024, Revenue Growth Continues

Sensor Tower’s annual 2025 State of Mobile report is here and it talks about the latest trends on mobile, highlighting that the mobile market is stronger than ever. According to the report, Google Play and App Store’s revenue for in-app purchases reached $150 billion which is a 13% YoY increase. There were 136 billion downloads made on Google Play and iOS YoY and the downloads have been between 135-140 billion since 2020. New smartphone users are slowing down but this doesn't mean that people are spending less time on their phones because consumers spent 4.2 trillion hours on different apps from Google Play and iOS, which makes 3.5 hours per day. The growth in total hours spent declined a bit from 7.7% YoY in 2023 to 5.8% YoY in 2024.

Downloads and time spent decreased in established markets but it didn't affect revenue growth. The revenue from in-app purchases on Google Play and iOS increased 13% YoY in 2024 and reached $150 billion and it was mostly due to non-gaming apps. There was a 4% YoY growth in gaming apps and 23% YoY growth in non-gaming apps. The market leading in IAP revenue in 2024 was US ($52 billion), with a 16% YoY increase. Europe saw a 24% YoY revenue growth in 2024. The in-app purchase revenue for mobile games grew 4% YoY to $81 billion in 2024 after seeing two years of decline. However, downloads declined 6% YoY to 49.6 billion.

AI apps are also doing good on Google Play and iOS and are already a billion dollar industry. AI Art generator and AI chatbot got $1.3 billion from in-app purchases with the US being 45% of the market for generative AI apps in terms of revenue, followed by 4% of the UK, 4% of Germany and 3% of Japan. More than 100 apps have added AI related terms to their apps names or tags which shows that AI technology has grown a lot and is expected to grow more in 2025.

Some other things that the report talks about is cryptocurrency apps are gaining transactions again after 2021 while social media apps are also seeing a quick growth in their revenue in 2024. Many of the apps on Google Play and iOS are also connecting mobile to in-person experiences like grocery store, outdoor health and restaurant apps.

Read next: Social Media’s 'News-Finds-Me' Mentality Fuels Misinformation Sharing
by Arooj Ahmed via Digital Information World

UK Cracks Down on Smugglers, Removes 8,000 Social Media Accounts Promoting Illegal Migration Services

The UK's National Crime Agency (NCA) has intensified efforts to disrupt people-smuggling networks operating on platforms like Meta, X, TikTok, and YouTube. By working closely with these companies, it achieved a 40% rise in account removals last year, taking down over 8,000 accounts promoting illegal migration services. This brings the three-year total to more than 16,500 takedowns.

Illegal posts included false claims of swift boat crossings to the UK, rewards for referring migrants, and sales of fake IDs. Smugglers also used these platforms to lure migrants with promises of transport across Europe, later switching to encrypted apps to avoid detection.

A robust social media action plan introduced in late 2021 has helped the NCA investigate and prosecute offenders. Preston-based Amanj Hasan Zada received a 17-year sentence in 2024 for sharing videos of smuggling success stories. Two others, Dilshad Shamo and Ali Khdir, were found guilty of advertising smuggling services and await sentencing.

The agency is ramping up resources to disrupt these networks, collaborating with platforms to identify and remove harmful content. Tough new UK laws, announced this month, include travel bans, social media restrictions, and phone-use limitations to tackle human smuggling.

In the US, Mexican cartels exploit social media for similar purposes. Cartel del Noreste (CDN) uses Facebook to recruit and coordinate operations, combining smuggling with violent criminal activities. Two high-ranking members were sentenced last year for their roles in leveraging social media to expand their network and evade authorities.

Image: NationalCrimeAgency / UK

Read next: Is DeepSeek the AI That Will Topple Google and OpenAI? Here's What You Need to Know
by Asim BN via Digital Information World

Is DeepSeek the AI That Will Topple Google and OpenAI? Here's What You Need to Know

In January 2025, DeepSeek, a Chinese AI chatbot, made waves in the global market. It quickly attracted attention, sparking debates about the future of AI and the shifting balance of technological dominance. While many have focused on its impressive capabilities, others have raised concerns about its underlying influences. DeepSeek signals a new direction in AI, one that challenges established norms and introduces new possibilities in both technology and geopolitics.

The Birth of DeepSeek

DeepSeek was founded in 2023 by Liang Wenfeng, a former quantitative finance expert. His leap into artificial intelligence was driven by a desire to build a chatbot that could combine human-like reasoning with practical problem-solving features. By January 2025, DeepSeek had launched its flagship product, the DeepSeek-R1 AI chatbot. Despite being in its early stages, the company’s smart development and bold approach have positioned it as a formidable contender in the AI space.

The Technology Behind DeepSeek

DeepSeek's technology relies on a mix of advance approaches to AI development. What sets it apart is how efficiently it uses resources. While companies like OpenAI and Google have poured billions into their AI systems, DeepSeek developed a competitive product with just under $6 million. This lean development has allowed DeepSeek to avoid the need for expensive hardware and massive data centers, which have been the norm in AI development.

The core of DeepSeek's model is its ability to function with minimal computing power while still delivering fast, accurate responses. Unlike its American counterparts, which require vast amounts of processing power and other resources, DeepSeek uses more efficient algorithms and less resource-intensive models. As a result, it’s able to provide similar performance at a fraction of the cost, making AI more accessible to companies and individuals alike.

Open-Source Revolution

One of DeepSeek's most significant moves is its decision to release the chatbot’s code as open-source software. Unlike most major AI companies that keep their code locked behind proprietary systems, DeepSeek has made its technology available to everyone under an MIT license. This is a big deal because it encourages the kind of collaboration and innovation that can drive accelerated technological advancement.

• Also read: Why Are AI Giants Betting Big on Washington, and What’s at Stake for the Future?

By making its source code open, DeepSeek is democratizing access to AI. Developers, businesses, and researchers can not only use the chatbot but also improve upon it, adapt it to different needs, and contribute to its ongoing development. This open-source approach is a stark contrast to the more closed-off models from major AI companies.

DeepSeek’s Impact on the AI Market

DeepSeek is already shaking up the AI market in ways that go beyond its technology. The cost-effective nature of the product is one of its most significant selling points. Businesses that previously couldn’t afford AI now have access to a powerful tool that can help them scale operations, automate tasks, and enhance decision-making processes.

Notably, by making AI technology affordable and accessible, DeepSeek is challenging the business models of major AI players. Companies like OpenAI have traditionally relied on subscription fees, enterprise contracts, and other monetization strategies. In contrast, DeepSeek's API service is priced much lower, making it an attractive option for smaller firms and independent developers. This shift in how AI services are priced could force larger companies to reconsider their pricing models or risk losing market share.

Political and Geopolitical Tensions

DeepSeek's rise isn't just about technology, it's also about geopolitics. As China continues to assert itself in the global tech landscape, DeepSeek’s success is seen as part of a broader strategy to challenge the dominance of U.S. companies in the AI industry. While the U.S. has been the driving force behind most AI innovations, China is quickly catching up, and DeepSeek is leading the charge.

• Related: Global IT Spending to Reach $5.61 Trillion in 2025, Driven by AI Investments

DeepSeek’s success is symbolic of China’s growing influence in the AI sector. The country has been investing heavily in AI research and development, and DeepSeek is reaping the benefits of this state-backed push. While the Chinese government has allowed the company to operate with relative autonomy, there are signs that this could change. Recent moves by the government to increase its investment in AI could signal greater involvement in DeepSeek's future.

This shift in power dynamics has already had consequences. For instance, Nvidia, a key supplier of chips used in AI models, saw its stock dropped in billions in response to DeepSeek's rise. Investors are now questioning whether the traditional methods of building AI, that is relying on high-end hardware and data centers, are sustainable in the face of more efficient, cost-effective alternatives.

Censorship and Ethical Dilemmas

One of the most significant criticisms of DeepSeek is its censorship practices. Like many Chinese tech companies, DeepSeek has limits on the type of content its AI can discuss. For example, when asked about the 1989 Tiananmen Square protests, DeepSeek’s chatbot refuses to answer, instead redirecting users to other topics. This self-censorship is a reflection of China’s strict policies on controlling the flow of information.

The issue of censorship raises important questions about the ethical implications of AI. While AI has the potential to facilitate open dialogue and promote free expression, it can also be used to suppress inconvenient truths. DeepSeek’s refusal to address sensitive historical events is a reminder of the risks that come with allowing government influence over AI technology.

Though DeepSeek’s approach to censorship mirrors that of other Chinese tech companies, there are signs that the company is not entirely subject to government control. Analysts believe that the Chinese government has largely stayed out of DeepSeek's operations, but this could change as the company grows and attracts more attention. The increasing investment in AI research by the Chinese government suggests that DeepSeek might eventually face greater scrutiny or even direct intervention.

Looking Ahead: What’s Next for DeepSeek?

As DeepSeek moves forward, its future is filled with both opportunities and challenges. The company has already proven that it can build a competitive product with limited resources, but scaling that success will require overcoming several hurdles. One of the biggest challenges will be ensuring that DeepSeek-R1 can handle more complex tasks without compromising its efficiency.

• Also read: ChatGPT Leads Global AI Search Trends, Google Gemini Follows Closely

DeepSeek’s open-source model will be crucial in this process. As more developers and companies adopt the technology, the chatbot will continue to evolve and improve. However, whether DeepSeek can maintain its momentum against the likes of OpenAI, Google, and other tech giants remains to be seen. If it can continue to innovate while keeping costs low, the company could redefine how AI is developed and used in the future.

Conclusion

DeepSeek represents a shift in how AI is built, accessed, and monetized. Its cost-effective, open-source model is reshaping the industry, making AI more accessible to businesses and developers of all sizes. At the same time, DeepSeek is a reminder that the future of AI will be shaped not just by technological advancements but also by the political and ethical decisions that govern its development.

While DeepSeek’s rise signals a new era in AI, it also raises important questions about how AI should be regulated, who controls it, and how it can be used to promote open dialogue and transparency. As the company continues to grow and improve, the world will be watching to see if it can challenge the established powers in the AI space and redefine the future of technology.

Have you tried DeepSeek yet? Share your experience with us in the comments or tag us on social media!

Image: DIW-Aigen

Read next:

• Did the App Store Create Hundreds of New Millionaires in 2024? Here’s the Proof!

• Social Media’s 'News-Finds-Me' Mentality Fuels Misinformation Sharing

• Which U.S. Cities Are Seeing the Fastest Rise in Digital Crime and Fraud Reports?
by Asim BN via Digital Information World

Which U.S. Cities Are Seeing the Fastest Rise in Digital Crime and Fraud Reports?

According to the FTC, 2.5 million Americans became victims of fraud in 2023, losing about $10 billion online. In response to this alarming trend, All About Cookies conducted a survey to identify which U.S. cities have the highest rates of online scam, based on fraud and identity theft reports. They gave scores from 1-100, with 100 being the worst score and it was also observed that most scams happen in the cities which have aging populations and warm climates. Miami was the most scammiest city in the US with score of 72.0. There were 1,775 average fraud reports per 100k population, with 162 spam call complaints within the same range.

Followed by Miami is Las Vegas, with 200 spam calls reported per 100k and 488 identity theft reports per 100k as well. The third scammiest city in the US is Orlando (1,602 fraud reports and 167 spam call complaints per 100k). Tampa has the fourth highest scam rate in the country with 1,623 fraud reports per 100k. Another city in the top five scammiest city in the US is Atlanta which had 1,988 fraud reports per 100k, with highest number of robocalls (4,133) received per person. There was a 29% decrease in identity theft reports in Atlanta and that's why it isn't ranked higher. Other top ten scammiest cities in the US include San Diego, Tallahassee, Tucson, Hartford and Dallas. Most of the cities with highest rates of scam and identity theft are from Florida.

All About Cookies also ranked the safest cities in America for digital crime, with Amarillo (18.6 score) topping the list. It also had a 21% decrease in identity theft reports which helped a lot in its ranking. Other safest cities in the US for digital media are Fort Wayne, Wichita, Nashville and Kansas City. Fort Wayne also saw a 14% decline in fraud reports while Kansas City has the lowest fraud reports and robocall rates in the country. On the other hand, Hartford saw a 49% increase in identity theft reports. Des Moines saw the second biggest increase in identity theft reports (40%) while Boston saw the third biggest increase (34%). Other cities in the US where identity theft is rising quickly are Omaha, Lincoln and Denver. Cities like Little Rock, Cincinnati, Providence and Portland are also seeing increases of 15%-13% in identity theft reports.

Read next: Survey Highlights $3,313 Average Loss Per Identity Theft Victim in the US
by Arooj Ahmed via Digital Information World

Why Are AI Giants Betting Big on Washington, and What’s at Stake for the Future?

According to OpenSecrets, many AI companies have significantly increased their lobbying efforts on federal AI issues. In 2023, 458 companies spent on AI lobbying while 648 companies spent on AI lobbying in 2024. This is a 141% YoY increase in AI lobbying between 2023 and 2024. Companies like Microsoft and OpenAI have ramped up their efforts to influence AI-related legislation. CREATE AI Act, which focuses on benchmarking AI systems in the US and Advancement and Reliability Act which aims at creating a government center for AI research was backed up by Microsoft and OpenAI respectively.

OpenAI has increased its lobbying expenditure from $260,000 in 2023 to $1.76 million in 2024 while Anthropic’s lobbying expenditure increased from $280,000 to $720,000 in 2024. A startup named Cohere has also increased its budget to $230,000 in 2024 from $70,000 in 2022. OpenAI, Cohere and Anthropic collectively spent $2.71 million on federal lobbying in 2024, which is a significant increase from $610,000 being spent on federal lobbying in 2023. It is still small if we compare it to what large tech industries spent on lobbying in 2023 ($61.5 million).

In 2024, domestic policy making was a mess with Congress considering about 90 AI-related bills in the first half of the year but Congress didn't take any actions and asked the states to act independently. Some of the actions which were taken because of that were Tennessee becoming the first state to protect voice artists from unauthorized AI cloning, Colorado adopting a risk based approach to AI policy and California enacting multiple AI safety bills. But no state was successful in implementing AI regulations as good as the EU's AI Act.

It is still unclear whether there will be more actions on AI legislation this year as compared to last year at a federal level. President Donal Trump has recently ordered federal agencies to suspend Biden-era AI policies, even the export rules on AI models. Anthropic has urged the federal government to implement targeted AI regulation while OpenAI has also called for some government action on AI as well as its development and infrastructure.

Image: DIW-Aigen

Read next:

• Meta AI Launches Document Editor to Enhance Writing and Editing Tasks

• Apple iOS 18 Adoption Hits 68%, Surpassing iOS 17 by 2% Amid Sluggish Generative AI Interest
by Arooj Ahmed via Digital Information World

Meta AI Launches Document Editor to Enhance Writing and Editing Tasks

Meta AI is offering its users a text-based document editor to help them in their writing tasks. If you want to write documents using AI assistance, head to the web client on Meta AI and open the document. You will have options just like any other document and Meta’s Llama 3.2 will help you make any kind of changes to your document. This tool is not on Meta AI yet and Meta hasn't shared any news about it either. But it is mentioned on Meta’s AI homepage that the feature is available on beta. Six weeks ago, a help page for this feature also appeared but there's no announcement about its public release yet.

You can also generate images on the document using Meta’s imagine feature with options about saving, copying and printing the document. However, the option to download the document in formats like DOCX is not yet available. Users can instruct Meta AI to make changes to the document and will also be able to take advantage of versioning features. If you do not like a change AI made, you can still go back to previous versions using the back arrow.

Many students use generative AI to cheat on their homework and writing tasks. When writing generated by Meta AI in the document editor was checked for AI, it was all detected as AI. This means that it cannot be used for writing homework without going undetected so using the rewrite feature will come in handy. Meta AI in the document editor can also be used for fixing spelling and grammatical mistakes.

H/T: Neowin

Read next: 

• Study Reveals Social Media Users Share News Without Reading, Driving Misinformation

• Survey Highlights $3,313 Average Loss Per Identity Theft Victim in the US
by Arooj Ahmed via Digital Information World