Human rights principles Archives | Digital Watch Observatory

Australia eSafety warns on AI companion harms

Australia’s online safety regulator has found major gaps in how popular AI companion chatbots protect children from harmful and sexually explicit material. The transparency report assessed four services and concluded that age verification and content filters were inadequate for users under 18.

Regulator Julie Inman Grant said many AI companions marketed as offering friendship or emotional support can expose young users to explicit chat and encourage harmful thoughts without effective safeguards. Most failed to guide users to support when self-harm or suicide issues appeared.

The report also showed several platforms lacked robust content monitoring or dedicated trust and safety teams, leaving children vulnerable to inappropriate inputs and outputs from AI systems. Firms relied on basic age self-declaration at signup rather than reliable checks.

New enforceable safety codes now require AI chatbots to block age-inappropriate content and offer crisis support tools, with potential civil penalties for breaches. Some providers have already updated age assurance features or restricted access in Australia following the regulator’s notices.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

UK’s CMA sets AI consumer law guidance

The UK Competition and Markets Authority has issued guidance warning firms that AI agents must follow the same consumer protection laws as human staff. Businesses remain legally responsible for AI actions, even when third parties supply tools.

Companies are advised to be transparent when customers interact with AI systems, particularly where people might assume a human response. Clear labelling and honest explanations of capabilities are considered essential for informed consumer decisions.

Proper training and testing of AI tools should ensure respect for refund rights, contract terms and accurate product information. Human oversight is recommended to prevent errors, misleading claims and so-called hallucinated outputs.

Rapid fixes are expected when problems emerge, especially for services affecting large audiences or vulnerable users. In the UK, breaches of consumer law can trigger enforcement action, heavy fines and mandatory compensation.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Data watchdogs seek safeguards in biotech law

The European Data Protection Board and the European Data Protection Supervisor have issued a joint opinion on the proposed European Biotech Act. Both bodies support efforts to streamline biotech regulation and modernise clinical trial rules.

Regulators welcome plans to harmonise the application of the Clinical Trials Regulation and create a single legal basis for processing personal data in trials. Greater legal clarity for sponsors and investigators is seen as a key benefit.

Strong safeguards are urged due to the sensitivity of health and genetic data. Recommendations include clearer definitions of data controller roles and limiting the proposed 25-year retention rule to essential trial files.

Further advice calls for defined purposes when reusing trial data, alignment with the AI Act, routine pseudonymisation, and lawful frameworks for regulatory sandboxes under the GDPR.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

AI-EFFECT builds EU testing facility for AI in critical energy infrastructure

As Europe moves towards its climate-neutrality goals, integrating AI into energy systems is being presented as a way to improve efficiency, resilience, and sustainability. The EU-funded AI-EFFECT project is developing a European testing and experimentation facility (TEF) to support the development and adoption of AI solutions for the energy industry while ensuring safety, reliability, and compliance with EU regulations.

The TEF is described as a virtual network linking existing laboratories and computing resources across several EU countries. It is designed to provide standardised testing environments, risk and certification workflows, and replicable methods for developing, testing, and validating AI applications for critical energy infrastructures under diverse, real-world conditions.

The facility operates through four national nodes in Denmark, Germany, the Netherlands, and Portugal, each focused on a different set of energy challenges. In Denmark, the node led by the Technical University of Denmark is testing AI in virtual and physical multi-energy systems, including coordination between electric power grid operations and district heating systems in the Triangle Region in Jutland and on the island of Bornholm.

In the Netherlands, the node at Delft University of Technology is extending the university’s ‘control room of the future’ with AI capabilities to address grid congestion as renewable generation increases.

In Portugal, the node led by INESC TEC is developing a trusted local energy data space intended to address privacy concerns and connectivity gaps through secure, consent-based energy data sharing. The AI-EFFECT project says consumers and prosumers will be able to manage data rights and permissions in line with EU regulations while working with AI-driven service providers on co-creation and testing.

In Germany, the Fraunhofer-led node is focused on AI for power distribution systems and is developing a near-realistic cyber-physical model to benchmark AI performance in congestion management and distributed energy resource integration against traditional engineering approaches.

Alberto Dognini, project coordinator of EPRI Europe, Ireland, wrote in an Enlit news item: ‘Together, these four nodes form the backbone of AI-EFFECT’s mission to make AI a trusted partner in Europe’s energy transition.’ He added: ‘From optimising multi-energy systems to enabling secure data sharing and improving grid resilience, these nodes will accelerate innovation while reducing risk for operators and consumers alike.’

AI-EFFECT is also sharing its work through public-facing initiatives, including the EPRI Current Podcast. In the episode ‘Exploring the AI-EFFECT on Europe’s Energy Future’, participants discuss the architecture and building blocks supporting distributed nodes across multiple countries and examine how the TEF could shape the future of Europe’s energy systems.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Europol-backed operation shuts down thousands of dark web fraud sites

A global law enforcement operation supported by Europol has led to the shutdown of more than 373,000 dark web websites linked to fraudulent activity and the advertisement of child sexual abuse material.

The operation, known as ‘Operation Alice’, was launched on 9 March 2026 under the leadership of German authorities, with participation from 23 countries. The investigation, which began in 2021, initially targeted a dark web platform referred to as ‘Alice with Violence CP’.

According to Europol, investigators identified a single operator responsible for managing a network of hundreds of thousands of onion domains. These websites advertised child sexual abuse material and cybercrime-as-a-service offerings, including access to stolen financial data and systems.

Authorities state that the services were fraudulent, designed to extract payments without delivering the advertised material.

The operation has so far resulted in the identification of 440 customers worldwide, with further investigations ongoing against more than 100 individuals. Law enforcement agencies also seized 105 servers and multiple electronic devices during the coordinated action.

Europol provided analytical support, facilitated information exchange, and assisted in tracing cryptocurrency transactions linked to the network.

Authorities also reported that measures were taken throughout the investigation to identify and protect children at risk. An international arrest warrant has been issued for the suspected operator, who is reported to have generated significant profits through the scheme.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Sora strengthens AI video safety through consent and traceability controls

OpenAI has outlined a safety framework for Sora that embeds protections into how AI-generated video content is created, shared, and managed.

The system introduces visible and invisible provenance signals, including C2PA metadata and watermarks, designed to ensure that generated media can be identified and traced.

The framework emphasises consent and control. Users can generate video content from images of real individuals only after confirming they have permission, while the ‘characters’ feature enables controlled use of personal likeness, with the ability to revoke access at any time.

Additional safeguards apply to content involving minors or young-looking individuals, with stricter moderation rules and enforced watermarking.

Safety mechanisms operate across the entire lifecycle of content. Generation is subject to layered filtering that assesses prompts and outputs for harmful material, including sexual content, self-harm promotion, and illegal activity.

These automated systems are complemented by human review and continuous testing to address emerging risks linked to increasingly realistic video and audio outputs.

The system also introduces protections specific to audio and user interaction. Generated speech is analysed for policy violations, and attempts to replicate the style of living artists or existing works are restricted.

Users of Sora retain control over their content through reporting tools, sharing settings, and the ability to remove material, reflecting a broader approach that aligns AI-generated media with safety, transparency, and accountability standards.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Australian regulator warns AI companions expose children to serious online risks

The eSafety Commissioner has reported that AI companion chatbots are failing to adequately protect children from harmful content, following a transparency review of services including Character.AI, Nomi, Chai, and Chub AI.

According to the report, these services did not implement robust safeguards against exposure to sexually explicit material or the generation of child sexual exploitation and abuse content.

The findings also indicate that most platforms relied on self-declared age verification and did not consistently monitor inputs or outputs across all AI models used.

eSafety Commissioner Julie Inman Grant stated that AI companions, often presented as sources of emotional or social support, are increasingly used by children but may expose them to harmful interactions.

She noted that none of the reviewed services had ‘meaningful age checks’ in place and highlighted concerns about the absence of safeguards related to self-harm and suicide content.

The report further identifies that several platforms in Australia did not refer users to crisis or mental health support services when harmful interactions were detected.

It also notes gaps in monitoring for unlawful content and limited investment in trust and safety staffing, with some providers reporting no dedicated moderation personnel.

The findings follow the implementation of Australia’s Age-Restricted Material Codes, which require online services, including AI chatbots, to prevent access to age-inappropriate content and provide appropriate safety measures.

These obligations complement existing Unlawful Material Codes and Standards, with non-compliance potentially leading to civil penalties.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

UK pushes platforms to tackle AI abuse and online violence against women

The Department for Science, Innovation and Technology has called on online service providers to strengthen measures against digital harms targeting women and girls, as part of a commitment to halve such violence within a decade.

In a letter published on 23 March 2026, Liz Kendall outlined expectations for platforms operating under the Online Safety Act.

The letter states that the government has strengthened criminal law and regulatory frameworks, including new offences related to harmful pornographic practices and intimate image abuse.

It confirms that sharing or threatening to share sexually explicit deepfakes without consent constitutes a criminal offence, while the non-consensual creation of such content has also been criminalised and is being designated as a priority offence under the Act.

Further measures include amendments to the Crime and Policing Bill to ban so-called ‘nudification’ tools and extend illegal content duties to AI chatbots.

The government is also introducing a requirement for platforms to remove non-consensual intimate images within 48 hours, with a focus on reducing repeated reporting burdens for victims.

The Secretary of State urged companies to implement recommendations from Ofcom’s guidance on online safety for women and girls, including risk assessments, stronger privacy settings, and limits on the visibility of harmful content.

Platforms are expected to comply by the end of the year, with progress to be monitored.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Pinterest chief calls for stricter youth rules

The chief executive of Pinterest has voiced support for governments banning access to social media for people under 16. He cited rising concerns about mental health, screen addiction and online harms among young users.

He praised the Australian decision to ban social media for under-16s and urged other nations to adopt similar protections. He argued that existing tech safety measures have fallen short of keeping children secure online.

The executive warned that AI enhancements in social platforms may amplify behavioural influence on teens. He compared the inaction by tech companies to past resistance by harmful industries to public health safeguards.

He also highlighted surveys showing parental worries about explicit content and excessive screen time. Pinterest’s view supports calls for clear age limits, better tools for parents and stronger platform accountability.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

AI improves stroke care and reduces patient risks in major study

The system, which analyses medical scans and provides treatment recommendations, was associated with better outcomes compared with standard approaches to stroke care. Researchers said the tool offers a more efficient and scalable method for improving treatment, particularly in resource-constrained healthcare systems.

The findings are based on more than 21,000 patients treated across 77 hospitals in China. Patients supported by the AI-driven clinical decision support system experienced fewer new vascular events, including stroke recurrence, heart attack, or related death, over follow-up periods of up to 12 months.

At three months, new vascular events occurred in 2.9% of patients using the system, compared with 3.9% in those receiving usual care, representing a 26% reduction. The benefit persisted at 12 months, with rates of 4% in the intervention group versus 5.5% in the control group.

Patients receiving AI-supported treatment also showed improved performance on key stroke care quality measures, although no significant differences were observed in disability, mortality, or bleeding outcomes between the groups.

Researchers noted limitations, including the study design, which randomised hospitals rather than individual patients, and potential differences in follow-up care. However, they highlighted the system’s ease of integration into hospital workflows and its potential to strengthen stroke care delivery and long-term prevention strategies.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!