Internet of things

Large language models mimic human object perception

Recent research shows that large multimodal language models (LLMs) can develop object representations strikingly similar to human cognition. By analysing how these AI models understand and organise concepts, scientists found patterns in the models that mirror neural activity in the human brain.

The study examined embeddings for 1,854 natural objects, derived from millions of text-image pairings. These embeddings capture relationships between objects and were compared with brain scan data from regions like EBA, PPA, RSC and FFA.

Researchers also discovered that multimodal training, which combines text and image data, enhances model’s ability to form these human-like concepts. Findings suggest that large language models can achieve more natural understanding of the world, offering potential improvements in human-AI interaction and future model design.

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

ChatGPT to exit WhatsApp after Meta policy change

OpenAI says ChatGPT will leave WhatsApp on 15 January 2026 after Meta’s new rules banning general-purpose AI chatbots on the platform. ChatGPT will remain available on iOS, Android, and the web, the company said.

Users are urged to link their WhatsApp number to a ChatGPT account to preserve history, as WhatsApp doesn’t support chat exports. OpenAI will also let users unlink their phone numbers after linking.

Until now, users could message ChatGPT on WhatsApp to ask questions, search the web, generate images, or talk to the assistant. Similar third-party bots offered comparable features.

Meta quietly updated WhatsApp’s business API to prohibit AI providers from accessing or using it, directly or indirectly. The change effectively forces ChatGPT, Perplexity, Luzia, Poke, and others to shut down their WhatsApp bots.

The move highlights platform risk for AI assistants and shifts demand toward native apps and web. Businesses relying on WhatsApp AI automations will need alternatives that comply with Meta’s policies.

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

AWS outage turned a mundane DNS slip into global chaos

Cloudflare’s boss summed up the mood after Monday’s chaos, relieved his firm wasn’t to blame as outages rippled across more than 1,000 companies. Snapchat, Reddit, Roblox, Fortnite, banks, and government portals faltered together, exposing how much of the web leans on Amazon Web Services.

AWS is the backbone for a vast slice of the internet, renting compute, storage, and databases so firms avoid running their own stacks. However, a mundane Domain Name System error in its Northern Virginia region scrambled routing, leaving services online yet unreachable as traffic lost its map.

Engineers call it a classic failure mode: ‘It’s always DNS.’ Misconfigurations, maintenance slips, or server faults can cascade quickly across shared platforms. AWS says teams moved to mitigate, but the episode showed how a small mistake at scale becomes a global headache in minutes.

Experts warned of concentration risk: when one hyperscaler stumbles, many fall. Yet few true alternatives exist at AWS’s scale beyond Microsoft Azure and Google Cloud, with smaller rivals from IBM to Alibaba, and fledgling European plays, far behind.

Calls for UK–EU cloud sovereignty are growing, but timelines and costs are steep. Monday’s outage is a reminder that resilience needs multi-region and multi-cloud designs, tested failovers, and clear incident comms, not just faith in a single provider.

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

Startup raises $9m to orchestrate Gulf digital infrastructure

Bilal Abu-Ghazaleh has launched 1001 AI, a London–Dubai startup building an AI-native operating system for critical MENA industries. The two-month-old firm raised $9m seed from CIV, General Catalyst and Lux Capital, with angels including Chris Ré, Amjad Masad and Amira Sajwani.

Target sectors include airports, ports, construction, and oil and gas, where 1001 AI sees billions in avoidable inefficiencies. Its engine ingests live operational data, models workflows and issues real-time directives, rerouting vehicles, reassigning crews and adjusting plans autonomously.

Abu-Ghazaleh brings scale-up experience from Hive AI and Scale AI, where he led GenAI operations and contributor networks. 1001 borrows a consulting-style rollout: embed with clients, co-develop the model, then standardise reusable patterns across similar operational flows.

Investors argue the Gulf is an ideal test bed given sovereign-backed AI ambitions and under-digitised, mission-critical infrastructure. Deena Shakir of Lux says the region is ripe for AI that optimises physical operations at scale, from flight turnarounds to cargo moves.

First deployments are slated for construction by year-end, with aviation and logistics to follow. The funding supports early pilots and hiring across engineering, operations and go-to-market, as 1001 aims to become the Gulf’s orchestration layer before expanding globally.

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

AWS outage shows the cost of cloud concentration

A single fault can bring down the modern web. During the outage on Monday, 20 October 2025, millions woke to broken apps, games, banking, and tools after database errors at Amazon Web Services rippled outward. When a shared backbone stumbles, the blast radius engulfs everything from chat to commerce.

The outage underscored cloud concentration risk. Roblox, Fortnite, Pokémon Go, Snapchat, and workplace staples like Slack and Monday.com stumbled together because many depend on the same region and data layer. Failover, throttling, and retries help, but simultaneous strain can swamp safeguards.

On Friday, 19 July 2024, a faulty CrowdStrike update crashed Windows machines worldwide, triggering blue screens that grounded flights, delayed surgeries, and froze point-of-sale systems. The fix was simple; recovery wasn’t. Friday patches gained a new cautionary tale.

Earlier shocks foreshadowed today’s scale. In 1997, a Network Solutions glitch briefly hobbled .com and .net. In 2018, malware in Alaska’s Matanuska-Susitna knocked services offline, sending a community of 100,000 back to paper. Each incident showed how mundane errors cascade into civic life.

Resilience now means multi-region designs, cross-cloud failovers, tested runbooks, rate-limit backstops, and graceful read-only modes. Add regulatory stress tests, clear incident comms, and sector drills with hospitals, airlines, and banks. The internet will keep breaking; our job is to make it bend.

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

Tailored pricing is here and personal data is the price signal

AI is quietly changing how prices are set online. Beyond demand-based shifts, companies increasingly tailor offers to individuals, using browsing history, purchase habits, device, and location to predict willingness to pay. Two shoppers may see different prices for the same product at the same moment.

Dynamic pricing raises or lowers prices for everyone as conditions change, such as school-holiday airfares or hotel rates during major events. Personalised pricing goes further by shaping offers for specific users, rewarding cart-abandoners with discounts while charging rarer shoppers a premium.

Platforms mine clicks, time on page, past purchases, and abandoned baskets to build profiles. Experiments show targeted discounts can lift sales while capping promo spend, proving engineered prices scale. The result: you may not see a ‘standard’ price, but one designed for you.

The risks are mounting. Income proxies such as postcode or device can entrench inequality, while hidden algorithms erode trust when buyers later find cheaper prices. Accountability is murky if tailored prices mislead, discriminate, or breach consumer protections without clear disclosure.

Regulators are moving. A competition watchdog in Australia has flagged transparency gaps, unfair trading risks, and the need for algorithmic disclosure. Businesses now face a twin test: deploy AI pricing with consent, explainability, and opt-outs, and prove it delivers value without crossing ethical lines.

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

AWS glitch triggers widespread outages across major apps

A major internet outage hit some of the world’s biggest apps and sites from about 9 a.m. CET Monday, with issues traced to Amazon Web Services. Tracking sites reported widespread failures across the US and beyond, disrupting consumer and enterprise services.

AWS cited ‘significant error rates’ in DynamoDB requests in the US-EAST-1 region, impacting additional services in Northern Virginia. Engineers are mitigating while investigating root cause, and some customers couldn’t create or update Support Cases.

Outages clustered around Virginia’s dense data-centre corridor but rippled globally. Impacted brands included Amazon, Google, Snapchat, Roblox, Fortnite, Canva, Coinbase, Slack, Signal, Vodafone and the UK tax authority HMRC.

Coinbase told users ‘all funds are safe’ as platforms struggled to authenticate, fetch data and serve content tied to affected back-ends. Third-party monitors noted elevated failure rates across APIs and app logins.

The incident underscores heavy reliance on hyperscale infrastructure and the blast radius when core data services falter. Full restoration and a formal post-mortem are pending from AWS.

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

Adult erotica tests OpenAI’s safety claims

OpenAI will loosen some ChatGPT rules, letting users make replies friendlier and allowing erotica for verified adults. Altman framed the shift as ‘treat adult users like adults’, tied to stricter age-gating. The move follows months of new guardrails against sycophancy and harmful dynamics.

The change arrives after reports of vulnerable users forming unhealthy attachments to earlier models. OpenAI has since launched GPT-5 with reduced sycophancy and behaviour routing, plus safeguards for minors and a mental-health council. Critics question whether evidence justifies loosening limits so soon.

Erotic role-play can boost engagement, raising concerns that at-risk users may stay online longer. Access will be restricted to verified adults via age prediction and, if contested, ID checks. That trade-off intensifies privacy tensions around document uploads and potential errors.

It is unclear whether permissive policies will extend to voice, image, or video features, or how regional laws will apply to them. OpenAI says it is not ‘usage-maxxing’ but balancing utility with safety. Observers note that ambitions to reach a billion users heighten moderation pressures.

Supporters cite overdue flexibility for consenting adults and more natural conversation. Opponents warn normalising intimate AI may outpace evidence on mental-health impacts. Age checks can fail, and vulnerable users may slip through without robust oversight.

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

An awards win for McAfee’s consumer-first AI defence

McAfee won ‘Best Use of AI in Cybersecurity’ at the 2025 A.I. Awards for its Scam Detector. The tool, which McAfee says is the first to automate deepfake, email, and text-scam detection, underscores a consumer-focused defence. The award recognises its bid to counter fast-evolving online fraud.

Scams are at record levels, with one in three US residents reporting victimisation and average losses of $1,500. Threats now range from fake job offers and text messages to AI-generated deepfakes, increasing the pressure on tools that can act in real time across channels.

McAfee’s Scam Detector uses advanced AI to analyse text, email, and video, blocking dangerous links and flagging deepfakes before they cause harm. It is included with core McAfee plans and available on PC, mobile, and web, positioning it as a default layer for everyday protection.

Adoption has been rapid, with the product crossing one million users in its first months, according to the company. Judges praised its proactive protection and emphasis on accuracy and trust, citing its potential to restore user confidence as AI-enabled deception becomes more sophisticated.

McAfee frames the award as validation of its responsible, consumer-first AI strategy. The company says it will expand Scam Detector’s capabilities while partnering with the wider ecosystem to keep users a step ahead of emerging threats, both online and offline.

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

A common EU layer for age verification without a single age limit

Denmark will push for EU-wide age-verification rules to avoid a patchwork of national systems. As Council presidency, Copenhagen prioritises child protection online while keeping flexibility on national age limits. The aim is coordination without a single ‘digital majority’ age.

Ministers plan to give the European Commission a clear mandate for interoperable, privacy-preserving tools. An updated blueprint is being piloted in five states and aligns with the EU Digital Identity Wallet, which is due by the end of 2026. Goal: seamless, cross-border checks with minimal data exposure.

Copenhagen’s domestic agenda moves in parallel with a proposed ban on under-15 social media use. The government will consult national parties and EU partners on the scope and enforcement. Talks in Horsens, Denmark, signalled support for stronger safeguards and EU-level verification.

The emerging compromise separates ‘how to verify’ at the EU level from ‘what age to set’ at the national level. Proponents argue this avoids fragmentation while respecting domestic choices; critics warn implementation must minimise privacy risks and platform dependency.

Next steps include expanding pilots, formalising the Commission’s mandate, and publishing impact assessments. Clear standards on data minimisation, parental consent, and appeals will be vital. Affordable compliance for SMEs and independent oversight can sustain public trust.

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