Online education Archives | Page 29 of 65 | Digital Watch Observatory

Concerns raised over Google AI Overviews and health advice

A Guardian investigation has found that Google’s AI Overviews have displayed false and misleading health information that could put people at risk of harm. The summaries, which appear at the top of search results, are generated using AI and are presented as reliable snapshots of key information.

The investigation identified multiple cases where Google’s AI summaries provided inaccurate medical advice. Examples included incorrect guidance for pancreatic cancer patients, misleading explanations of liver blood test results, and false information about women’s cancer screening.

Health experts warned that such errors could lead people to dismiss symptoms, delay treatment, or follow harmful advice. Some charities said the summaries lacked essential context and could mislead users during moments of anxiety or crisis.

Concerns were also raised about inconsistencies, with the same health queries producing different AI-generated answers at different times. Experts said this variability undermines trust and increases the risk that misinformation will influence health decisions.

Google said most AI Overviews are accurate and helpful, and that the company continually improves quality, particularly for health-related topics. It said action is taken when summaries misinterpret content or lack appropriate context.

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

Universities in Ireland urged to rethink assessments amid AI concerns

Face-to-face interviews and oral verification could become a routine part of third-level assessments under new recommendations aimed at addressing the improper use of AI. Institutions are being encouraged to redesign assessment methods to ensure student work is authentic.

The proposals are set out in new guidelines published by the Higher Education Authority (HEA) of Ireland, which regulates universities and other third-level institutions. The report argues that assessment systems must evolve to reflect the growing use of generative AI in education.

While encouraging institutions to embrace AI’s potential, the report stresses the need to ensure students are demonstrating genuine learning. Academics have raised concerns that AI-generated assignments are increasingly difficult to distinguish from original student work.

To address this, the report recommends redesigning assessments to prioritise student authorship and human judgement. Suggested measures include oral verification, process-based learning, and, where appropriate, a renewed reliance on written exams conducted without technology.

The authors also caution against relying on AI detection tools, arguing that integrity processes should be based on dialogue and evidence. They call for clearer policies, staff and student training, and safeguards around data use and equitable access to AI tools.

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

New οffline AI note app promises privacy without subscriptions

Growing concern over data privacy and subscription fatigue has led an independent developer to create WitNote, an AI note-taking tool that runs entirely offline.

The software allows users to process notes locally on Windows and macOS rather than relying on cloud-based services where personal information may be exposed.

WitNote supports lightweight language models such as Qwen2.5-0.5B that can run with limited storage requirements. Users may also connect to external models through API keys if preferred.

Core functions include rewriting, summarising and extending content, while a WYSIWYG Markdown editor provides a familiar workflow without network delays, instead of relying on web-based interfaces.

Another key feature is direct integration with Obsidian Markdown files, allowing notes to be imported instantly and managed in one place.

The developer says the project remains a work in progress but commits to ongoing updates and user-driven improvements, even joining Apple’s developer programme personally to support smoother installation.

For users seeking AI assistance while protecting privacy and avoiding monthly fees, WitNote positions itself as an appealing offline alternative that keeps full control of data on the local machine.

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

Belgium’s influencers seek clarity through a new certification scheme

The booming influencer economy of Belgium is colliding with an advertising rulebook that many creators say belongs to another era.

Different obligations across federal, regional and local authorities mean that wording acceptable in one region may trigger a reprimand in another. Some influencers have even faced large fines for administrative breaches such as failing to publish business details on their profiles.

In response, the Influencer Marketing Alliance in Belgium has launched a certification scheme designed to help creators navigate the legal maze instead of risking unintentional violations.

Influencers complete an online course on advertising and consumer law and must pass a final exam before being listed in a public registry monitored by the Jury for Ethical Practices.

Major brands, including L’Oréal and Coca-Cola, already prefer to collaborate with certified creators to ensure compliance and credibility.

Not everyone is convinced.

Some Belgian influencers argue that certification adds more bureaucracy at a time when they already struggle to understand overlapping rules. Others see value as a structured reminder that content creators remain legally responsible for commercial communication shared with followers.

The alliance is also pushing lawmakers to involve influencers more closely when drafting future rules, including taxation and safeguards for child creators.

Consumer groups such as BEUC support clearer definitions and obligations under the forthcoming EU Digital Fairness Act, arguing that influencer advertising should follow the same standards as other media instead of remaining in a grey zone.

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

Best AI dictation tools for faster speech-to-text in 2026

AI dictation reached maturity during the years after many attempts of patchy performance and frustrating inaccuracies.

Advances in speech-to-text engines and large language models now allow modern dictation tools to recognise everyday speech more reliably while keeping enough context to format sentences automatically instead of producing raw transcripts that require heavy editing.

Several leading apps have emerged with different strengths. Wispr Flow focuses on flexibility with style options and custom vocabulary, while Willow blends automation with privacy by storing transcripts locally.

Monologue also prioritises privacy by allowing users to download the model and run transcription entirely on their own machines. Superwhisper caters for power users by supporting multiple downloadable models and transcription from audio or video files.

Other tools take different approaches. VoiceTypr offers an offline-first design with lifetime licensing, Aqua promotes speed and phrase-based shortcuts, Handy provides a simple free open source starting point, and Typeless gives one of the most generous free allowances while promising strong data protection.

Each reflects a wider trend where developers try to balance convenience, privacy, control and affordability.

Users now benefit from cleaner, more natural-sounding transcripts instead of the rigid audio typing tools of previous years. AI dictation has become faster, more accurate and far more usable for everyday note-taking, messaging and work tasks.

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

Best AI chatbot for maths accuracy revealed in new benchmark

AI tools are increasingly used for simple everyday calculations, yet a new benchmark suggests accuracy remains unreliable.

The ORCA study tested five major chatbots across 500 real-world maths prompts and found that users still face roughly a 40 percent chance of receiving the wrong answer.

Gemini from Google recorded the highest score at 63 percent, with xAI’s Grok almost level at 62.8 percent. DeepSeek followed with 52 percent, while ChatGPT scored 49.4 percent, and Claude placed last at 45.2 percent.

Performance varied sharply across subjects, with maths and conversion tasks producing the best results, but physics questions dragged scores down to an average accuracy below 40 percent.

Researchers identified most errors as sloppy calculations or rounding mistakes, rather than deeper failures to understand the problem. Finance and economics questions highlighted the widest gaps between the models, while DeepSeek struggled most in biology and chemistry, with barely one correct answer in ten.

Users are advised to double-check results whenever accuracy is crucial. A calculator or a verified source is still advised instead of relying entirely on an AI chatbot for numerical certainty.

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

Visually impaired gamers call for more accessible games

Many visually impaired gamers find mainstream video games difficult due to limited accessibility features. Support groups enable players to share tips, recommend titles, and connect with others who face similar challenges.

Audio and text‑based mobile games are popular, yet console and PC titles often lack voiceovers or screen reader support. Adjustable visual presets could make mainstream games more accessible for partially sighted players.

UK industry bodies acknowledge progress, but barriers remain for millions of visually impaired players. Communities offer social support and provide feedback to developers to improve games and make them inclusive.

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

AI cheating drives ACCA to halt online exams

The Association of Chartered Certified Accountants (ACCA) has announced it will largely end remote examinations in the UK from March 2026, requiring students to sit tests in person unless exceptional circumstances apply.

The decision aims to address a surge in cheating, particularly facilitated by AI tools.

Remote testing was introduced during the Covid-19 pandemic to allow students to continue qualifying when in-person exams were impossible. The ACCA said online assessments have now become too difficult to monitor effectively, despite efforts to strengthen safeguards against misconduct.

Investigations show cheating has impacted major auditing firms, including the ‘big four’ and other top companies. High-profile cases, such as EY’s $100m (£74m) settlement in the US, highlight the risks posed by compromised professional examinations.

While other accounting bodies, including the Institute of Chartered Accountants in England and Wales, continue to allow some online exams, the ACCA has indicated that high-stakes assessments must now be conducted in person to maintain credibility and integrity.

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

OpenAI strengthened ChatGPT Atlas with new protections against prompt injection attacks

Protecting AI agents from manipulation has become a top priority for OpenAI after rolling out a major security upgrade to ChatGPT Atlas.

The browser-based agent now includes stronger safeguards against prompt injection attacks, where hidden instructions inside emails, documents or webpages attempt to redirect the agent’s behaviour instead of following the user’s commands.

Prompt injection poses a unique risk because Atlas can carry out actions that a person would normally perform inside a browser. A malicious email or webpage could attempt to trigger data exposure, unauthorised transactions or file deletion.

Criminals exploit the fact that agents process large volumes of content across an almost unlimited online surface.

OpenAI has developed an automated red-team framework that uses reinforcement learning to simulate sophisticated attackers.

When fresh attack patterns are discovered, the models behind Atlas are retrained so that resistance is built into the agent rather than added afterwards. Monitoring and safety controls are also updated using real attack traces.

These new protections are already live for all Atlas users. OpenAI advises people to limit logged-in access where possible, check confirmation prompts carefully and give agents well-scoped tasks instead of broad instructions.

The company argues that proactive defence is essential as agentic AI becomes more capable and widely deployed.

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

AI is changing how Europeans work and learn

Generative AI has become an everyday tool across Europe, with millions using platforms such as ChatGPT, Gemini and Grok for personal, work, and educational purposes. Eurostat data shows that around a third of people aged 16–74 tried AI tools at least once in 2025.

Adoption varies widely across the continent. Norway leads with 56 percent of the population using AI, while Turkey records only 17 percent.

Within the EU, Denmark tops usage at 48 percent, and Romania lags at 18 percent. Northern and digitally advanced countries dominate, while southern, central-eastern, and Balkan nations show lower engagement.

Researchers attribute this to general digital literacy, internet use, and familiarity with technology rather than government policy alone. AI tools are used more for personal purposes than for work.

Across the EU, 25 percent use AI for personal tasks, compared with 15 percent for professional applications.

Usage in education is even lower, with only 9 percent employing AI in formal learning, peaking at 21 percent in Sweden and Switzerland and dropping to just 1 percent in Hungary.

Experts stress that while access is essential, understanding how to apply AI effectively remains a key barrier. Countries with strong digital foundations adopt AI more, while limited awareness and skills restrict use, emphasising the need for AI literacy and infrastructure.

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