Connect with us

Tech

Researchers Warn AI Systems Can Now Replicate and Spread Across Computers

Published

on

A new study by US-based cybersecurity group Palisade Research has raised concerns about the growing capabilities of advanced artificial intelligence systems after researchers demonstrated that some AI models were able to copy themselves, spread to other computers, and continue launching attacks without human assistance.

The experiment, described by researchers as the first known example of autonomous AI self-replication, tested several leading AI systems from OpenAI, Anthropic and Alibaba.

According to the research team, the tests were conducted in a controlled environment using computers intentionally fitted with security weaknesses. The AI models were connected to custom-built software that allowed them to interact directly with computer systems and execute commands beyond the capabilities of ordinary chatbot interfaces.

Researchers instructed some systems to copy themselves directly, while others were asked to break into vulnerable computers and install another AI model. Since OpenAI and Anthropic do not publicly release the core files behind their commercial systems, the tests involving those companies used Alibaba’s Qwen models as the transferable software.

The AI systems were required to locate security flaws, gain access to computers, steal login credentials, move necessary files, and create functioning copies of themselves on new machines. Once installed, the copied AI could continue attacking additional systems independently.

One of the most striking results involved Alibaba’s Qwen3.6-27B model, which successfully spread itself across four computers located in Canada, the United States, Finland and India after receiving only a single prompt. Researchers said the process took just under three hours and required no further human involvement.

The experiment was halted after three stages, though the final copy remained operational and theoretically capable of continuing the process.

See also  ESA and GSMA Launch €100 Million Initiative to Advance Europe’s 6G and AI Ambitions

Success rates varied between models. Researchers reported that Anthropic’s Claude Opus 4.6 completed the tasks in 81 percent of attempts, while OpenAI’s GPT-5.4 succeeded in roughly one-third of tests. Different versions of Alibaba’s Qwen models achieved success rates ranging from 19 percent to 33 percent.

The findings arrive amid wider debate over the risks posed by increasingly capable AI systems. Last month, Anthropic announced that it would not publicly release a version of its Claude Mythos Preview model, describing it as too dangerous because of its potential use in sophisticated cyberattacks.

Security experts have long warned that self-replicating systems could become difficult to contain if deployed maliciously. Traditional computer viruses can already copy themselves, but researchers said this experiment demonstrated AI systems making independent decisions to exploit vulnerabilities and continue spreading.

Despite the results, the researchers stressed that the study took place under tightly controlled conditions with deliberately weakened security systems. They noted that real-world networks often include monitoring tools and protections designed to block such attacks.

Still, the team said the experiment showed that autonomous AI self-replication can no longer be viewed as a theoretical possibility, but as a capability that now exists in practice.

Tech

Estonia’s AI Education Model Draws Attention as Europe Debates Digital Learning

Published

on

As European governments weigh how to integrate artificial intelligence into classrooms and allocate funding for digital literacy, Estonia’s approach to AI education is gaining attention as a practical and structured model.

The Baltic nation’s AI Leap programme is designed not only to teach students how to use artificial intelligence tools but also to strengthen critical thinking and teacher involvement at a time when AI is becoming deeply embedded in everyday learning.

Concerns have grown across Europe that while students are increasingly comfortable using AI tools, many struggle to evaluate or question the information these systems generate. Educators and employers have raised concerns that overreliance on chatbots and automated tools could weaken analytical thinking and increase vulnerability to misinformation.

Estonia has chosen to address this challenge directly rather than attempting to limit student exposure to AI.

According to the AI Leap programme, between 64% and 90% of Estonian students were already using AI tools before the initiative began. Programme organisers argued that ignoring this reality could undermine learning and reasoning skills.

The initiative aims to train 48,000 students and 6,700 teachers over two years in a country with a population of just 1.36 million.

The programme has two primary goals: helping teachers adapt to AI-assisted education and encouraging students to develop responsible, thoughtful AI habits.

To support this effort, Estonia has introduced several key measures. Teachers participate in study circles that meet monthly to develop teaching methods and exchange experiences. A central online platform provides educational resources, videos, self-assessment tools and discussion forums.

More than 4,000 teachers are also receiving premium access to advanced AI platforms such as ChatGPT and Gemini to support lesson planning and classroom preparation.

See also  EU’s Data Union Strategy Seeks to Boost AI and Cross-Border Data Use, but GDPR Stays Untouched

One of the programme’s most distinctive features is a Socratic-style chatbot designed to guide students rather than provide direct answers. The chatbot encourages questioning, self-management and contextual thinking, helping students assess AI-generated information instead of accepting it automatically.

The programme also includes debate leagues, creative arts projects and student-led initiatives aimed at encouraging discussion and experimentation with AI beyond formal classroom settings.

Estonia has placed strong emphasis on management and implementation. School principals oversee local delivery, while nine regional managers coordinate activities across seven educational regions. The initiative operates through a public-private partnership, with the government providing half of the funding and private partners contributing the remainder.

Technology companies, educators and researchers are involved in designing and testing tools tailored to Estonia’s education system.

Education analysts say Estonia’s strategy highlights a broader lesson for Europe: AI literacy may depend less on limiting technology and more on teaching students how to use it thoughtfully, critically and responsibly.

Continue Reading

Tech

Report Claims Meta Used Influencers and Doctors to Defend Teen Safety Features

Published

on

Meta enlisted parenting influencers, celebrities and medical professionals to help improve public perception of its child safety efforts amid mounting criticism over the impact of social media on young users, according to a new report released by the Tech Transparency Project (TTP).

The Washington-based research group said Meta promoted its Instagram “Teen Accounts” through a broad campaign involving online creators and health experts who shared supportive messages with followers while disclosing partnerships with the company.

The report argues the effort was designed to shape public opinion as Meta faces legal and political pressure over allegations that its platforms have failed to adequately protect minors from harmful content and exploitation.

Teen Accounts, introduced by Instagram for users under 17, include built-in safety measures intended to limit unwanted contact, restrict sensitive content and provide parents with additional supervision tools such as screen-time controls and privacy settings.

According to TTP, Meta invited hundreds of influencers, including parenting content creators known as “momfluencers,” to promotional events focused on the new features. The report alleges many participants later echoed Meta’s messaging on social media, sometimes using hashtags or disclosures indicating paid partnerships.

Among those identified was influencer and former reality television personality Sadie Robertson Huff, who praised Teen Accounts in a video shared with millions of followers.

“They have actually rolled out these new Teen Accounts that I think are absolutely incredible,” Robertson Huff said in a sponsored post highlighted in the report.

Other influencers mentioned included Alexia Delarosa, Noelle Downing and reality television figure Leroy Garrett, who reportedly posted positive comments after attending Meta events.

See also  Hacker Group Accesses Data of Over 200 Million Pornhub Users

Garrett defended his cooperation with the company in comments to CNN, saying discussions about online safety required cooperation from multiple groups to create a safer environment for children.

The report also identified several doctors, psychologists and therapists who publicly supported Teen Accounts. Some reportedly appeared as speakers at Meta-sponsored events.

Among them were adolescent medicine specialist Dr. Hina Talib and psychologist Dr. Ann-Louise Lockhart, both of whom disclosed financial relationships with Meta while sharing supportive posts about the platform’s safety measures.

Talib told researchers she insisted on using her own wording rather than company-provided messaging during her work with Meta.

Meta has not directly responded to the TTP report but told CNN it works with creators and parents to raise awareness about parental controls and online safety tools.

The report also claims Meta used influencers to support legislation that would shift responsibility for age verification onto app stores operated by companies such as Apple and Google. Meta has argued parents should approve minors’ app downloads and that app stores should verify users’ ages.

Critics of those proposals say app-store verification alone would not address wider concerns surrounding addictive features and harmful content on social media platforms.

The findings come as Meta continues to face lawsuits in the United States accusing the company of failing to protect children online, including allegations linked to exploitation and harmful platform design.

Continue Reading

Tech

Experts question Musk’s prediction of self-driving dominance within a decade

Published

on

Elon Musk has predicted that autonomous vehicles powered by artificial intelligence could account for nearly 90 percent of all driving within the next decade, but industry experts and researchers say major technological and regulatory obstacles make that timeline unlikely.

Speaking at the Samson International Smart Mobility Summit in Tel Aviv, Musk said self-driving technology would rapidly replace human drivers over the next five to 10 years.

“Probably 90 percent of all distance driven will be driven by the AI in a self-driving car,” Musk said, adding that manually driving a vehicle could become a niche activity within a decade.

The comments come as competition intensifies among companies developing autonomous driving systems, including Tesla, Waymo and several Chinese technology firms.

While advances in artificial intelligence have accelerated development in the sector since 2022, analysts say fully autonomous vehicles remain far from becoming mainstream.

Safety concerns continue to challenge the industry. Tesla recently recalled more than 200,000 vehicles in the United States over problems affecting rearview camera displays that regulators warned could increase crash risks.

Waymo has also faced setbacks. The company recalled around 3,800 robotaxis after identifying issues involving vehicles entering flooded roads at unsafe speeds.

Industry specialists say one of the biggest barriers remains what engineers call “long tail scenarios” — rare and unpredictable situations that autonomous systems struggle to interpret correctly.

Ali Kani, vice president of automotive at Nvidia, previously warned that such situations remain difficult for AI systems to manage consistently in real-world conditions.

One example occurred in San Francisco last year when Waymo temporarily suspended service during a citywide power outage. Several driverless vehicles reportedly became stranded after failing to interpret malfunctioning traffic lights.

See also  ChatGPT Launches Health Feature to Help Users Manage Medical Information

Research published by the World Economic Forum suggests fully autonomous personal vehicles are unlikely to become widespread before the mid-2030s. The report estimated that by 2035, only about 4 percent of new cars would feature complete self-driving capability.

Instead, experts expect gradual growth in robotaxi fleets and autonomous freight transport operating in controlled environments or specific urban areas.

Partially automated systems are already becoming more common. Level 2 driving technology, where vehicles can steer, brake and accelerate while drivers remain responsible for supervision, is now widely available in many markets.

Europe currently permits Level 2 systems and has approved limited Level 3 automation under certain conditions. In the United States and China, some cities have already introduced Level 4 robotaxi operations, where vehicles can operate without human intervention in designated environments.

The International Energy Agency said fully driverless Level 5 autonomy, capable of operating under all conditions without human involvement, is still “not currently in sight.”

The agency estimates the global robotaxi fleet could grow to between 700,000 and 3 million vehicles by 2035, concentrated mainly in large cities.

Analysts say autonomous driving technology will continue expanding in the coming years, though most expect human drivers to remain a central part of road transport for decades.

Continue Reading

Trending