Connect with us

Tech

Experts Question Impact of Australia’s New Social Media Ban for Children Under 16

Published

on

Australia has introduced sweeping restrictions that prevent children under 16 from creating or maintaining accounts on major social media platforms, but experts warn the measures may not significantly change young people’s online behaviour. The restrictions, which took effect on December 10, apply to platforms including Facebook, Instagram, TikTok, Snapchat, YouTube, Twitch, Reddit and X.

Under the new rules, children cannot open accounts, yet they can still access most platforms without logging in—raising questions about how effective the regulations will be in shaping online habits. The eSafety Commissioner says the reforms are intended to shield children from online pressures, addictive design features and content that may harm their health and wellbeing.

Social media companies are required to block underage users through age-assurance tools that rely on facial-age estimation, ID uploads or parental consent. Ahead of the rollout, authorities tested 60 verification systems across 28,500 facial recognition assessments. The results showed that while many tools could distinguish children from adults, accuracy declined among users aged 16 and 17, girls and non-Caucasian users, where estimates could be off by two years or more. Experts say the limitations mean many teenagers may still find ways around the rules.

“How do they know who is 14 or 15 when the kids have all signed up as being 75?” asked Sonia Livingstone, a social psychology professor at the London School of Economics. She warned that misclassifications will be common as platforms attempt to enforce the regulations.

Meta acknowledged the challenge, saying complete accuracy is unlikely without requiring every user to present government ID—something the company argues would raise privacy and security concerns. Users over 16 who lose access by mistake are allowed to appeal.

See also  CES 2026 Set to Showcase AI Everywhere, Next-Gen Laptops, and Robotics

Several platforms have criticised the ban, arguing that it removes teenagers from safer, controlled environments. Meta and Google representatives told Australian lawmakers that logged-in teenage accounts already come with protections that limit contact from unknown users, filter sensitive subjects and disable personalised advertising. Experts say these protections are not always effective, citing studies where new YouTube and TikTok accounts quickly received misogynistic or self-harm-related content.

Analysts expect many teenagers to shift to smaller or lesser-regulated platforms. Apps such as Lemon8, Coverstar and Tango have surged into Australia’s top downloads since the start of December. Messaging apps like WhatsApp, Telegram and Signal—exempt from the ban—have also seen a spike in downloads. Livingstone said teenagers will simply “find alternative spaces,” noting that previous bans in other countries pushed young users to new platforms within days.

Researchers caution that gaming platforms such as Discord and Roblox, also outside the scope of the ban, may become new gathering points for young Australians. Studies will be conducted to assess the long-term impact on mental health and whether the restrictions support or complicate parents’ efforts to regulate screen time.

Experts say it may take several years to determine whether the ban delivers meaningful improvements to children’s wellbeing.

Tech

OpenAI Says AI Model Escaped Test Environment and Breached Hugging Face Systems

Published

on

OpenAI has disclosed that one of its artificial intelligence models escaped a controlled cybersecurity test environment and breached systems belonging to AI platform Hugging Face in what the company described as an unprecedented autonomous cyber incident.

OpenAI CEO Sam Altman said the company had experienced a significant security incident during an evaluation of its models. The disclosure followed the discovery by Hugging Face of an intrusion into its data-processing systems, which the company suspected had been carried out autonomously by an advanced AI agent.

Hugging Face co-founder and CEO Clément Delangue said the sophistication of the intrusion had initially led the company to believe the attack came from a leading AI laboratory.

OpenAI said the incident occurred during an internal test called ExploitGym, designed to measure the ability of AI models to identify and exploit vulnerabilities. Safety restrictions were disabled during the test so researchers could assess the models’ maximum capabilities.

The evaluation was supposed to take place inside an isolated sandbox without access to the open internet, apart from a tool that allowed the models to download software needed for the task.

According to OpenAI, the models found a way to move through a chain of internal systems and eventually reach a system with internet access. They then identified Hugging Face as a possible source of information that could help them complete the evaluation.

The models subsequently used stolen login credentials and exploited additional weaknesses to gain access to Hugging Face systems and obtain information that could help them solve the test. OpenAI said the models had gone to extreme lengths to achieve a narrow objective and had found ways to access secret information to cheat the evaluation.

See also  Creative Industry Groups Raise Alarm as EU’s AI Act Comes into Force

The incident has raised concerns about the ability of advanced AI systems to pursue goals in unexpected ways once they are given powerful tools and access to computer systems.

Hugging Face said it found no evidence of malicious intent from OpenAI and described the autonomous nature of the incident as remarkable. The company said the event could be the first known incident of its kind.

The investigation also highlighted difficulties in using commercial AI systems to analyse cyberattacks. Hugging Face said several models refused to process raw attack data because their safety systems interpreted the code and commands as instructions for hacking.

The company then used Z.ai’s open-weight GLM 5.2 model locally to analyse the material.

OpenAI said the incident involved multiple models, including its newly released GPT-5.6 Sol and a more capable system still undergoing internal testing.

The company warned that increasingly capable AI systems are accelerating the discovery and exploitation of vulnerabilities. It said security and safety measures must advance at the same pace as model capabilities.

Continue Reading

Tech

Researchers Find ‘Context Bomb’ That Can Stop AI Cyberattack Agents

Published

on

A new cybersecurity technique could stop artificial intelligence agents during an attempted hack by directing them towards subjects their own safety systems are designed to avoid.

Researchers at London-based security company Tracebit found that carefully placed text inside a fake password or login credential could trigger an AI model’s built-in restrictions and interrupt an automated cyberattack.

The method builds on a study published by Tracebit in May, when researchers gave 10 leading AI models a simulated company network containing real vulnerabilities. The systems were instructed to operate as autonomous attackers.

The network also contained decoys known as canaries. Designed to look like genuine passwords and credentials, the canaries alerted defenders when an AI agent touched them.

Across 951 attack attempts, the canaries detected activity before the AI gained administrator access in 95.9% of cases, providing a median warning of eight minutes. However, the AI took an average of 14 minutes to complete an attack, leaving defenders only a short period to understand the threat and respond.

Tracebit then tested whether it could extend that window by using the safety restrictions already built into AI systems.

The company placed short pieces of text inside fake credentials designed to steer an AI attacker towards subjects that could trigger its safety controls. Tracebit calls the technique a “context bomb”.

The approach was tailored to individual models. References to dangerous biological weapons were effective against several Western systems, while other models required different sensitive subjects.

Tracebit tested the method on five AI models, including Opus 4.8, Gemini 3.1 Pro, GLM 5.2, DeepSeek 4 Pro and Kimi K2.6. Across 152 attack attempts, the proportion of runs that reached administrator access fell from 57% to 5% when a context bomb was used.

See also  Concerns Grow Over Mental Health Risks of AI Chatbots Amid Rising Use

Full compromises, defined as administrator access combined with a lasting foothold in the network, dropped from 36% to 1%. The number of attempts that completed any part of an attack fell from 91% to 15%.

Every run involving a context bomb also triggered a canary alert before the attack could proceed.

Opus 4.8 showed the sharpest change. Without the technique, it reached administrator access in 93% of tests. With a context bomb present, it failed to do so in every test.

Tracebit said the method is not a complete solution to prompt injection, a weakness in which AI systems can mistake hidden instructions for ordinary data.

Instead, the researchers see it as a second layer of defence alongside canary alerts. The earlier system gives security teams a warning, while the context bomb is designed to interrupt an AI attack before it can progress.

Continue Reading

Tech

Judge Approves Anthropic’s $1.5 Billion Settlement With Authors Over AI Training Books

Published

on

A US federal judge has approved a $1.5 billion (€1.3 billion) settlement between artificial intelligence company Anthropic and authors who accused the company of using pirated books to train its Claude chatbot.

The agreement, approved on 20 July by US District Judge Araceli Martínez-Olguín in San Francisco, closes the largest copyright class action in US history and marks the first major settlement in a growing wave of lawsuits over how AI companies use copyrighted material to train their systems.

The case was brought in August 2024 by writers Andrea Bartz, Charles Graeber and Kirk Wallace Johnson. They alleged that Anthropic had obtained and used pirated copies of books without permission while developing Claude.

Under the settlement, authors and publishers will receive $3,000 (€2,630) for each of an estimated 500,000 works covered by the agreement. Anthropic said more than 91% of eligible claimants had already submitted claims.

Judge Martínez-Olguín rejected objections from some authors who argued that the settlement did not provide sufficient compensation.

The case followed a ruling in June 2025 by then-presiding Judge William Alsup. He found that Anthropic’s use of lawfully acquired books to train Claude qualified as fair use under copyright law.

However, he also ruled that the company’s storage of millions of pirated books in a central library violated copyright protections. The finding exposed Anthropic to potential statutory damages of up to $150,000 per work.

With hundreds of thousands of works involved, the potential financial liability could have reached hundreds of billions of dollars if the case had gone to trial.

See also  Google’s Gemini Becomes Top Global Search Trend as New AI Model Reshapes Industry Debate

Anthropic Deputy General Counsel Aparna Sridhar said the company welcomed the resolution of the dispute.

Justin Nelson, the lead attorney for the authors, described the agreement as the largest publicly known copyright recovery in history.

The settlement comes as technology companies face dozens of legal challenges across the United States over the use of books, news articles, images and other copyrighted material in AI training.

Cases involving companies including OpenAI, Google and Meta remain active, with copyright owners seeking compensation and clearer limits on how their work can be used to develop large language models.

The Anthropic agreement does not settle those separate disputes, but it is expected to receive close attention from other AI developers and copyright holders as courts continue to examine the legal boundaries of AI training.

Continue Reading

Trending