Artificial intelligence

DeepSeek dominates AI crypto trading challenge

Chinese AI model DeepSeek V3.1 has outperformed its global competitors in a real-market cryptocurrency trading challenge, earning over 10 per cent profit in just a few days.

The experiment, named Alpha Arena, was launched by US research firm Nof1 to test the investing skills of leading LLMs.

Each participating AI was given US$10,000 to trade in six cryptocurrency perpetual contracts, including bitcoin and solana, on the decentralised exchange Hyperliquid. By Tuesday afternoon, DeepSeek V3.1 led the field, while OpenAI’s GPT-5 trailed behind with a loss of nearly 40 per cent.

The competition highlights the growing potential of AI models to make autonomous financial decisions in real markets.

It also underscores the rivalry between Chinese and American AI developers as they push to demonstrate their models’ adaptability beyond traditional text-based tasks.

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

OpenAI launches ChatGPT Atlas web browser

OpenAI has launched ChatGPT Atlas, a web browser built around ChatGPT to help users work and explore online more efficiently. The browser lets ChatGPT operate directly on webpages, using past conversations and browsing context to assist with tasks without copying and pasting.

Early testers say it streamlines research, study, and productivity by providing instant AI support alongside the content they are viewing.

Atlas introduces browser memories, letting ChatGPT recall context from visited sites to improve responses and automate tasks. Users stay in control, with the ability to view, archive, or delete memories. 

Agent mode allows ChatGPT to perform tasks such as researching, summarising, or planning events while browsing. Safety is a priority, with safeguards to prevent unauthorised actions and options to operate in logged-out mode.

The browser is available worldwide on macOS for Free, Plus, Pro, and Go users, with Windows, iOS, and Android support coming soon. OpenAI plans to add multi-profile support, better developer tools, and improved app discoverability, advancing an agent-driven web experience with seamless AI integration.

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

MIT unveils SEAL, a self-improving AI model

Researchers at the Massachusetts Institute of Technology (MIT) have unveiled SEAL, a new AI model capable of improving its own performance without human intervention. The framework allows the model to generate its own training data and fine-tuning instructions, enabling it to learn new tasks autonomously.

The model employs reinforcement learning, a method in which it tests different strategies, evaluates their effectiveness, and adjusts its internal processes accordingly. This allows SEAL to refine its capabilities and increase accuracy over time.

In trials, SEAL outperformed GPT-4.1 by learning from the data it generated independently. The results demonstrate the potential of self-improving AI systems to reduce reliance on manually curated datasets and human-led fine-tuning.

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

Scouts can now earn AI and cybersecurity badges

In the United States, Scouting America, formerly known as the Boy Scouts, has introduced two new merit badges in AI and cybersecurity. The badges give scouts the opportunity to explore modern technology and understand its applications, while the organisation continues to adapt its programs to a digital era. Scouting America has around a million members and offers hundreds of merit badges across a wide range of skills.

The AI badge challenges scouts to examine AI’s effects on daily life, study deepfakes, and complete projects that demonstrate AI concepts. The cybersecurity badge teaches practical tools to stay safe online, emphasises ethical behaviour, and introduces scouts to a career field with thousands of unfilled positions.

Earlier this year, Scouting America launched Scoutly, an AI-powered chatbot designed to answer questions about the organisation and its merit badges. The initiative is part of Scouting America’s broader effort to modernise its programs and prepare young people for opportunities in an increasingly digital world.

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

Kenya leads the way in AI skilling across Africa

Kenya’s AI Skilling Initiative (AINSI) is offering valuable insights for African countries aiming to build digital capabilities. With AI projected to create 230 million digital jobs across Africa by 2030, coordinated investment in skills development is vital to unlock this potential.

Despite growing ambition, fragmented efforts and uneven progress continue to limit impact.

Government leadership plays a central role in building national AI capacity. Kenya’s Regional Centre of Competence for Digital and AI Skilling has trained thousands of public servants through structured bootcamps and online programmes.

Standardising credentials and aligning training with industry needs are crucial to ensure skilling efforts translate into meaningful employment.

Industry and the informal economy are key to scaling transformation. Partnerships with KEPSA and MESH are training entrepreneurs and SMEs in AI and cybersecurity while tackling affordability, connectivity, and data access challenges.

Education initiatives, from K–12 to universities and technical institutions, are embedding AI training into curricula to prepare future generations.

Civil society collaboration further broadens access, with community-based programmes reaching gig workers and underserved groups. Kenya’s approach shows how inclusive, cross-sector frameworks can scale digital skills and support Africa’s AI-driven growth.

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

YouTube launches likeness detection to protect creators from AI misuse

YouTube has expanded its AI safeguards with a new likeness detection system that identifies AI-generated videos imitating creators’ faces or voices. The tool is now available to eligible members of the YouTube Partner Program after a limited pilot phase.

Creators can review detected videos and request their removal under YouTube’s privacy rules or submit copyright claims.

YouTube said the feature aims to protect users from having their image used to promote products or spread misinformation without consent.

The onboarding process requires identity verification through a short selfie video and photo ID. Creators can opt out at any time, with scanning ending within a day of deactivation.

YouTube has backed recent legislative efforts, such as the NO FAKES Act in the US, which targets deceptive AI replicas. The move highlights growing industry concern over deepfake misuse and the protection of digital identity.

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

Meta strengthens protection for older adults against online scams

The US giant, Meta, has intensified its campaign against online scams targeting older adults, marking Cybersecurity Awareness Month with new safety tools and global partnerships.

Additionally, Meta said it had detected and disrupted nearly eight million fraudulent accounts on Facebook and Instagram since January, many linked to organised scam centres operating across Asia and the Middle East.

The social media giant is joining the National Elder Fraud Coordination Center in the US, alongside partners including Google, Microsoft and Walmart, to strengthen investigations into large-scale fraud operations.

It is also collaborating with law enforcement and research groups such as Graphika to identify scams involving fake customer service pages, fraudulent financial recovery services and deceptive home renovation schemes.

Meta continues to roll out product updates to improve online safety. WhatsApp now warns users when they share screens with unknown contacts, while Messenger is testing AI-powered scam detection that alerts users to suspicious messages.

Across Facebook, Instagram and WhatsApp, users can activate passkeys and complete a Security Checkup to reinforce account protection.

The company has also partnered with organisations worldwide to raise scam awareness among older adults, from digital literacy workshops in Bangkok to influencer-led safety campaigns across Europe and India.

These efforts form part of Meta’s ongoing drive to protect users through a mix of education, advanced technology and cross-industry cooperation.

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

Most EU workers now rely on digital tools and AI

A new EU study finds that 90% of workers rely on digital tools, while nearly a third use AI-powered chatbots in their daily work. The JRC and European Commission surveyed over 70,000 workers across all EU Member States between 2024 and 2025.

The findings show that AI is most commonly used for writing and translation tasks, followed by data processing and image generation. Adoption rates are particularly high in Northern and Central Europe, especially in office-based sectors.

Alongside this digital transformation, workplace monitoring is becoming increasingly widespread, with 37% of EU workers reporting that their working hours are tracked and 36% that their entry and exit times are monitored.

Algorithmic management, where digital systems allocate tasks or assess performance automatically, now affects about a quarter of EU workers. The study also identifies a growing ‘platformisation’ trend, categorising employees based on their exposure to digital monitoring and algorithmic control.

Workers facing full or physical platformisation often report higher stress levels and reduced autonomy, while informational platformisation appears to have milder effects, particularly for remote workers.

Researchers urge EU policymakers to curb digital oversight risks while promoting fair and responsible innovation. The findings support EU initiatives like the Quality Jobs Roadmap and efforts to regulate algorithmic management.

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

Dutch watchdog warns AI chatbots threaten election integrity

The Dutch data authority warns AI chatbots are biased and unreliable for voting advice ahead of national elections. An AP investigation found chatbots often steered users to the same two parties, ignoring their actual preferences.

In over half of the tests, the bots suggested either Geert Wilders’ far-right Freedom Party (PVV) or the leftwing GroenLinks-PvdA led by Frans Timmermans. Other parties, such as the centre-right CDA, were rarely mentioned even when users’ answers closely matched their platforms.

AP deputy head Monique Verdier said that voters were being steered towards parties that did not necessarily reflect their political views, warning that this undermines the integrity of free and fair elections.

The report comes ahead of the 29 October election, where the PVV currently leads the polls. However, the race remains tight, with GroenLinks-PvdA and CDA still in contention and many voters undecided.

Although the AP noted that the bias was not intentional, it attributed the problem to the way AI chatbots function, highlighting the risks of relying on opaque systems for democratic decisions.

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

Teachers become intelligence coaches in AI-driven learning

AI is reshaping education, pushing teachers to act as intelligence coaches and co-creators instead of traditional instructors.

Experts at an international conference, hosted in Greece, to celebrate Athens College’s centennial, discussed how AI personalises learning and demands a redefined teaching role.

Bill McDiarmid, professor emeritus at the University of North Carolina, said educators must now ask students where they find their information and why they trust it.

Similarly, Yong Zhao of the University of Kansas highlighted that AI enables individualised learning, allowing every student to achieve their full potential.

Speakers agreed AI should serve as a supportive partner, not a replacement, helping schools prepare students for an active role in shaping their futures.

The event, held under Greek President Konstantinos Tasoulas’ auspices, also urged caution when experimenting with AI on minors due to potential long-term risks.

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