Interdisciplinary approaches

Grok incident renews scrutiny of generative AI safety

Elon Musk’s Grok chatbot has triggered international backlash after generating sexualised images of women and girls in response to user prompts on X, raising renewed concerns over AI safeguards and platform accountability.

The images, some depicting minors in minimal clothing, circulated publicly before being removed. Grok later acknowledged failures in its own safeguards, stating that child sexual abuse material is illegal and prohibited, while xAI initially offered no public explanation.

European officials reacted swiftly. French ministers referred the matter to prosecutors, calling the output illegal, while campaigners in the UK argued the incident exposed delays in enforcing laws against AI-generated intimate images.

In contrast, US lawmakers largely stayed silent despite xAI holding a major defence contract. Musk did not directly address the controversy; instead, posting unrelated content as criticism mounted on the platform.

The episode has intensified debate over whether current AI governance frameworks are sufficient to prevent harm, particularly when generative systems operate at scale with limited real-time oversight.

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

ChatGPT reaches 40 million daily users for health advice

More than 40 million people worldwide now use ChatGPT daily for health-related advice, according to OpenAI.

Over 5 percent of all messages sent to the chatbot relate to healthcare, with three in five US adults reporting use in the past three months. Many interactions occur outside clinic hours, highlighting the demand for AI guidance in navigating complex medical systems.

Users primarily turn to AI to check symptoms, understand medical terms, and explore treatment options.

OpenAI emphasises that ChatGPT helps patients gain agency over their health, particularly in rural areas where hospitals and specialised services are scarce.

The technology also supports healthcare professionals by reducing administrative burdens and providing timely information.

Despite growing adoption, regulatory oversight remains limited. Some US states have attempted to regulate AI in healthcare, and lawsuits have emerged over cases where AI-generated advice has caused harm.

OpenAI argues that ChatGPT supplements rather than replaces medical services, helping patients interpret information, prepare for care, and navigate gaps in access.

Healthcare workers are also increasingly using AI. Surveys show that two in five US professionals, including nurses and pharmacists, use generative AI weekly to draft notes, summarise research, and streamline workflows.

OpenAI plans to release healthcare policy recommendations to guide the responsible adoption of AI in clinical settings.

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

AI-designed sensors open new paths for early cancer detection

MIT and Microsoft researchers have developed AI-designed molecular sensors to detect cancer in its earliest stages. By coating nanoparticles with peptides targeted by cancer-linked enzymes, the sensors produce signals detectable through simple urine tests, potentially even at home.

The AI system, named CleaveNet, generates peptide sequences that are efficiently and selectively cleaved by specific proteases, enzymes overactive in cancer cells. The approach enables faster, more precise detection and can help identify a tumour’s type and location.

CleaveNet, trained on 20,000+ peptide-protease interactions, has designed novel peptides for enzymes like MMP13 that cancer cells use to metastasise. The system may cut the number of peptides needed for diagnostics and reveal key biological pathways.

Researchers plan an at-home kit to detect 30 cancers, with peptides also usable for targeted therapies. The work is part of an ARPA-H-funded initiative and highlights the potential of AI to accelerate early cancer detection and treatment.

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

Healthcare systems face mounting risk from CrazyHunter ransomware

CrazyHunter ransomware has emerged as a growing threat to healthcare organisations, with repeated attacks targeting hospitals and medical service providers. The campaign focuses on critical healthcare infrastructure, raising concerns about service disruption and the exposure of sensitive patient data.

The malware is developed in Go and demonstrates a high level of technical maturity. Attackers gain initial access by exploiting weak Active Directory credentials, then use Group Policy Objects to distribute the ransomware rapidly across compromised networks.

Healthcare institutions in Taiwan have been among the most affected, with multiple confirmed incidents reported by security researchers. The pattern suggests a targeted campaign rather than opportunistic attacks, increasing pressure on regional healthcare providers to strengthen defences.

Once deployed, CrazyHunter turns off security tools and encrypts files to conceal its activity. Analysts note the use of extensive evasion techniques, including memory-based execution and redundant encryption methods, to ensure the delivery of the payload.

CrazyHunter employs a hybrid encryption scheme that combines ChaCha20 and elliptic curve cryptography, utilising partial file encryption to expedite the impact. Encrypted files receive a ‘.Hunter’ extension, with recovery dependent on the attackers’ private keys, reinforcing the pressure to pay ransoms.

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

AI assistant and cheaper autonomy headline Ford’s CES 2026 announcements

Ford has unveiled plans for an AI assistant that will launch in its smartphone app in early 2026 before expanding to in-vehicle systems in 2027. The announcement was made at the 2026 Consumer Electronics Show, alongside a preview of a next-generation BlueCruise driver assistance system.

The AI assistant will be hosted on Google Cloud and built using existing large language models, with access to vehicle-specific data. Ford said this will allow users to ask both general questions, such as vehicle capacity, and real-time queries, including oil life and maintenance status.

Ford plans to introduce the assistant first through its redesigned mobile app, with native integration into vehicles scheduled for 2027. The company has not yet specified which models will receive the in-car version first, but said the rollout would expand gradually across its lineup.

Alongside the AI assistant, the vehicle manufacturer previewed an updated version of its BlueCruise system, which it claims will be more affordable to produce and more capable. The new system is expected to debut in 2027 on the first electric vehicle built on Ford’s low-cost Universal Electric Vehicle platform.

Ford said the next-generation BlueCruise could support eyes-off driving by 2028 and enable point-to-point autonomous driving under driver supervision. As with similar systems from other automakers, drivers will still be required to remain ready to take control at any time.

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

Meet the voice-first AI companion with personality

Portola has launched Tolan, a voice-first AI companion that learns from ongoing conversations through personalised, animated characters. Tolan is designed for open-ended dialogue, making voice interactions more natural and engaging than standard text-based AI.

Built around memory and character design, the platform uses real-time context reconstruction to maintain personality and track shifting topics. Each turn, the system retrieves user memories, persona traits, and conversation tone, enabling coherent, adaptive responses.

GPT‑5.1 has improved latency, steerability, and consistency, reducing memory recall errors by 30% and boosting next-day retention over 20%.

Tolan’s architecture combines fast vector-based memory, dynamic emotional adjustment, and layered persona scaffolds. Sub-second responses and context rebuilding help the AI handle topic changes, maintain tone, and feel more human-like.

Since February 2025, Tolan has gained over 200,000 monthly users, earning a 4.8-star rating on the App Store. Future plans focus on multimodal voice agents integrating vision, context, and enhanced steerability to expand the boundaries of interactive AI.

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

AI and optical imaging transform thyroid cancer surgery

Thyroid cancer, the most common endocrine malignancy, poses challenges for surgeons trying to remove tumours while preserving healthy tissue.

Fine-needle aspiration and pathology are accurate but slow, providing no real-time guidance and sometimes causing unnecessary or incomplete surgeries. Dynamic Optical Contrast Imaging (DOCI) uses cells’ natural light to quickly distinguish healthy tissue from cancer.

The technique captures 23 optical channels from freshly excised tissue, creating detailed spectral maps without dyes or contrast agents. These optical signatures allow for rapid, label-free tissue analysis.

Researchers at Duke University and UCLA combined DOCI with AI to improve accuracy in classification and localisation. A two-stage machine-learning approach first categorised tissue as healthy or cancerous, including common and aggressive thyroid cancer subtypes.

Deep-learning models then produced tumour probability maps, pinpointing cancerous regions with minimal false positives.

Although initial studies focused on post-excision tissue, the technology could soon offer surgeons real-time guidance in the operating room. By combining optical imaging with AI, DOCI may reduce unnecessary surgery, preserve healthy tissue, and improve outcomes for thyroid cancer patients.

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

New UK cyber strategy focuses on trust in online public services

The UK government has announced new measures to strengthen the security and resilience of online public services as more interactions with the state move online. Ministers say public confidence is essential as citizens increasingly rely on digital systems for everyday services.

Backed by more than £210 million, the UK Government Cyber Action Plan outlines how cyber defences and digital resilience will be improved across the public sector. A new Government Cyber Unit will coordinate risk identification, incident response, and action on complex threats spanning multiple departments.

The plan underpins wider efforts to digitise public services, including benefits applications, tax payments, and healthcare access. Officials argue that secure systems can reduce bureaucracy and improve efficiency, but only if users trust that their data is protected.

The announcement coincides with parliamentary debate on the Cyber Security and Resilience Bill, which sets clearer expectations for companies supplying services to the government. The legislation is intended to strengthen cyber resilience across critical supply chains.

Ministers also highlighted new steps to address software supply chain risks, including a Software Security Ambassador Scheme promoting basic security practices. The government says stronger cyber resilience is essential to protect public services and maintain public trust.

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

ChatGPT Health offers personalised health support

OpenAI has launched ChatGPT Health, a secure platform linking users’ health information with ChatGPT’s intelligence. The platform supports, rather than replaces, medical care, helping users understand test results, prepare for appointments, and manage their wellness.

ChatGPT Health allows users to safely connect medical records and apps such as Apple Health, Function, and MyFitnessPal. All data is stored in a separate Health space with encryption and enhanced privacy to keep sensitive information secure.

Conversations in Health are not used to train OpenAI’s models.

The platform was developed with input from more than 260 physicians worldwide, ensuring guidance is accurate, clinically relevant, and prioritises safety.

HealthBench, a physician-informed evaluation framework, helps measure quality, clarity, and appropriate escalation in responses, supporting users in making informed decisions about their health.

ChatGPT Health is initially available outside the EEA, Switzerland, and the UK, with wider access coming in the coming weeks. Users can sign up for a waitlist and begin connecting records and wellness apps to receive personalised, context-aware health insights.

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

AI model turns sleep data into early disease predictions

Stanford Medicine researchers have developed an AI model that can analyse a single night of sleep to predict long-term disease risk. Known as SleepFM, the system uses physiological signals recorded during overnight sleep studies to identify early indicators of future health conditions.

The model was trained on nearly 600,000 hours of polysomnography data from 65,000 participants. Polysomnography captures brain activity, heart rhythms, breathing patterns, eye movements, and muscle signals, creating one of the most data-rich assessments used in medicine.

SleepFM was designed as a foundation model that learns how multiple biological signals interact during sleep. By reconstructing missing data streams, the system identifies patterns across different physiological systems rather than analysing signals in isolation.

After training, the model matched or outperformed existing tools in standard sleep data assessments, including sleep stage classification and sleep apnoea severity. Researchers then linked sleep data with long-term health records to evaluate its ability to predict future disease onset.

The model demonstrated strong predictive performance across 130 conditions, encompassing various diseases, including cancers, cardiovascular disease, and neurological disorders. Researchers say the findings position sleep data as an early warning signal, while further work will focus on interpretation and real-world clinical use.

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