Content Policy

Microsoft study flags 40 jobs highly vulnerable to AI automation

Microsoft Research released a comprehensive AI impact assessment, ranking 80 occupations by exposure to generative AI tools such as Copilot and ChatGPT. Roles heavily involved in language, writing, client communication, and routine digital tasks showed the highest AI overlap. Notable examples include translators, historians, customer service agents, political scientists, and data scientists.

By contrast, jobs requiring hands-on work, empathy, real-time physical or emotional engagement, such as nurses, phlebotomists, construction trades, embalmers, and housekeeping staff, were classified as low risk under current AI capabilities. Experts suggest that these kinds of positions remain essential because they involve physical presence, human interaction, and complex real-time decision making.

Although certain professions scored high for AI exposure, Microsoft and independent analysts emphasise that most jobs won’t disappear entirely. Instead, generative AI tools are expected to augment workflows, creating hybrid roles where human judgement and oversight remain critical, especially in sectors such as financial services, healthcare, and creative industries.

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

OpenAI pulls searchable chats from ChatGPT

OpenAI has removed a feature that allowed users to make their ChatGPT conversations publicly searchable, following backlash over accidental exposure of sensitive content.

Dane Stuckey, OpenAI’s CISO, confirmed the rollback on Thursday, describing it as a short-lived experiment meant to help users find helpful conversations. However, he acknowledged that the feature posed privacy risks.

‘Ultimately, we think this feature introduced too many opportunities for folks to accidentally share things they didn’t intend to,’ Stuckey wrote in a post on X. He added that OpenAI is working to remove any indexed content from search engines.

The move came swiftly after Fast Company and privacy advocate Luiza Jarovsky reported that some shared conversations were appearing in Google search results.

Jarovsky posted examples on X, noting that even though the chats were anonymised, users were unknowingly revealing personal experiences, including harassment and mental health struggles.

To activate the feature, users had to tick a box allowing their chat to be discoverable. While the process required active steps, critics warned that some users might opt in without fully understanding the consequences. Stuckey said the rollback will be complete by Friday morning.

The incident adds to growing concerns around AI and user privacy, particularly as conversational platforms like ChatGPT become more embedded in everyday life.

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

UK universities urged to act fast on AI teaching

UK universities risk losing their competitive edge unless they adopt a clear, forward-looking approach to ΑΙ in teaching. Falling enrolments, limited funding, and outdated digital systems have exposed a lack of AI literacy across many institutions.

As AI skills become essential for today’s workforce, employers increasingly expect graduates to be confident users rather than passive observers.

Many universities continue relying on legacy technology rather than exploring the full potential of modern learning platforms. AI tools can enhance teaching by adapting to individual student needs and helping educators identify learning gaps.

However, few staff have received adequate training, and many universities lack the resources or structure to embed AI into day-to-day teaching effectively.

To close the growing gap between education and the workplace, universities must explore flexible short courses and microcredentials that develop workplace-ready skills.

Introducing ethical standards and data transparency from the start will ensure AI is used responsibly without weakening academic integrity.

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

As Meta AI grows smarter on its own, critics warn of regulatory gaps

While OpenAI’s ChatGPT and Google’s Gemini dominate headlines, Meta’s AI is making quieter, but arguably more unsettling, progress. According to CEO Mark Zuckerberg, Meta’s AI is advancing rapidly and, crucially, learning to improve without external input.

In a blog post titled ‘Personal Superintelligence’, Zuckerberg claimed that Meta AI is becoming increasingly powerful through self-directed development. While he described current gains as modest, he emphasised that the trend is both real and significant.

Zuckerberg framed this as part of a broader mission to build AI that acts as a ‘personal superintelligence’, a tool that empowers individuals and becomes widely accessible. However, critics argue this narrative masks a deeper concern: AI systems that can evolve autonomously, outside human guidance or scrutiny.

The concept of self-improving AI is not new. Researchers have previously built systems capable of learning from other models or user interactions. What’s different now is the speed, scale and opacity of these developments, particularly within big tech companies operating with minimal public oversight.

The progress comes amid weak regulation. While governments like the Biden administration have issued AI action plans, experts say they lack the strength to keep up. Meanwhile, AI is rapidly spreading across everyday services, from healthcare and education to biometric verification.

Recent examples include Google’s behavioural age-estimation tools for teens, illustrating how AI is already making high-stakes decisions. As AI systems become more capable, questions arise: How much data will they access? Who controls them? And can the public meaningfully influence their design?

Zuckerberg struck an optimistic tone, framing Meta’s AI as democratic and empowering. However, that may obscure the risks of AI outpacing oversight, as some tech leaders warn of existential threats while others focus on commercial gains.

The lack of transparency worsens the problem. If Meta’s AI is already showing signs of self-improvement, are similar developments happening in other frontier models, such as GPT or Gemini? Without independent oversight, the public has no clear way to know—and even less ability to intervene.

Until enforceable global regulations are in place, society is left to trust that private firms will self-regulate, even as they compete in a high-stakes race for dominance. That’s a risky gamble when the technology itself is changing faster than we can respond.

As Meta AI evolves with little fanfare, the silence may be more ominous than reassuring. AI’s future may arrive before we are prepared to manage its consequences, and by then, it might be too late to shape it on our terms.

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

EU AI Act oversight and fines begin this August

A new phase of the EU AI Act takes effect on 2 August, requiring member states to appoint oversight authorities and enforce penalties. While the legislation has been in force for a year, this marks the beginning of real scrutiny for AI providers across Europe.

Under the new provisions, countries must notify the European Commission of which market surveillance authorities will monitor compliance. But many are expected to miss the deadline. Experts warn that without well-resourced and competent regulators, the risks to rights and safety could grow.

The complexity is significant. Member states must align enforcement with other regulations, such as the GDPR and Digital Services Act, raising concerns regarding legal fragmentation and inconsistent application. Some fear a repeat of the patchy enforcement seen under data protection laws.

Companies that violate the EU AI Act could face fines of up to €35 million or 7% of global turnover. Smaller firms may face reduced penalties, but enforcement will vary by country.

Rules regarding general-purpose AI models such as ChatGPT, Gemini, and Grok also take effect. A voluntary Code of Practice introduced in July aims to guide compliance, but only some firms, such as Google and OpenAI, have agreed to sign. Meta has refused, arguing the rules stifle innovation.

Existing AI tools have until 2027 to comply fully, but any launched after 2 August must meet the new requirements immediately. With implementation now underway, the AI Act is shifting from legislation to enforcement.

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

Scattered Spider cyberattacks set to intensify, warn FBI and CISA

The cybercriminal group known as Scattered Spider is expected to intensify its attacks in the coming weeks, according to a joint warning issued by the FBI, CISA, and cybersecurity agencies in Canada, the UK and Australia.

These warnings highlight the group’s increasingly sophisticated methods, including impersonating employees to bypass IT support and hijack multi-factor authentication processes.

Instead of relying on old techniques, the hackers now deploy stealthy tools like RattyRAT and DragonForce ransomware, particularly targeting VMware ESXi servers.

Their attacks combine social engineering with SIM swapping and phishing, enabling them to exfiltrate sensitive data before locking systems and demanding payment — a tactic known as double extortion.

Scattered Spider, also referred to as Okta Tempest, is reportedly creating fake online identities and infiltrating internal communication channels like Slack and Microsoft Teams. In some cases, they have even joined incident response calls to gain insight into how companies are reacting.

Security agencies urge organisations to adopt phishing-resistant multi-factor authentication, audit remote access software, monitor unusual logins and behaviours, and ensure offline encrypted backups are maintained.

More incidents are expected, as the group continues refining its strategies instead of slowing down.

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

TikTok adopts crowd‑sourced verification tool to combat misinformation

TikTok has rolled out Footnotes in the United States, its crowd‑sourced debunking initiative to supplement existing misinformation controls.

Vetted contributors will write and rate explanatory notes beneath videos flagged as misleading or ambiguous. If a note earns broad support, it becomes visible to all US users.

The system uses a ‘bridging‑based’ ranking framework to encourage agreement between users with differing viewpoints, making the process more robust and reducing partisan bias. Initially launched as a pilot, the platform has already enlisted nearly 80,000 eligible US users.

Footnotes complements TikTok’s integrity setup, including automated detection, human moderation, and partnerships with fact‑checking groups like AFP. Platform leaders note that effectiveness improves as contributors engage more across various topics.

Past research shows comparable crowd‑sourced systems often struggle to publish most submissions, with fewer than 10% of Notes appearing publicly on other platforms. Concerns remain over the system’s scalability and potential misuse.

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

Google rolls out AI age detection to protect teen users

In a move aimed at enhancing online protections for minors, Google has started rolling out a machine learning-based age estimation system for signed-in users in the United States.

The new system uses AI to identify users who are likely under the age of 18, with the goal of providing age-appropriate digital experiences and strengthening privacy safeguards.

Initially deployed to a small number of users, the system is part of Google’s broader initiative to align its platforms with the evolving needs of children and teenagers growing up in a digitally saturated world.

‘Children today are growing up with technology, not growing into it like previous generations. So we’re working directly with experts and educators to help you set boundaries and use technology in a way that’s right for your family,’ the company explained in a statement.

The system builds on changes first previewed earlier this year and reflects Google’s ongoing efforts to comply with regulatory expectations and public demand for better youth safety online.

Once a user is flagged by the AI as likely underage, Google will introduce a range of restrictions—most notably in advertising, content recommendation, and data usage.

According to the company, users identified as minors will have personalised advertising disabled and will be shielded from ad categories deemed sensitive. These protections will be enforced across Google’s entire advertising ecosystem, including AdSense, AdMob, and Ad Manager.

The company’s publishing partners were informed via email this week that no action will be required on their part, as the changes will be implemented automatically.

Google’s blog post titled ‘Ensuring a safer online experience for US kids and teens’ explains that its machine learning model estimates age based on behavioural signals, such as search history and video viewing patterns.

If a user is mistakenly flagged or wishes to confirm their age, Google will offer verification tools, including the option to upload a government-issued ID or submit a selfie.

The company stressed that the system is designed to respect user privacy and does not involve collecting new types of data. Instead, it aims to build a privacy-preserving infrastructure that supports responsible content delivery while minimising third-party data sharing.

Beyond advertising, the new protections extend into other parts of the user experience. For those flagged as minors, Google will disable Timeline location tracking in Google Maps and also add digital well-being features on YouTube, such as break reminders and bedtime prompts.

Google will also tweak recommendation algorithms to avoid promoting repetitive content on YouTube, and restrict access to adult-rated applications in the Play Store for flagged minors.

The initiative is not Google’s first foray into child safety technology. The company already offers Family Link for parental controls and YouTube Kids as a tailored platform for younger audiences.

However, the deployment of automated age estimation reflects a more systemic approach, using AI to enforce real-time, scalable safety measures. Google maintains that these updates are part of a long-term investment in user safety, digital literacy, and curating age-appropriate content.

Similar initiatives have already been tested in international markets, and the company announces it will closely monitor the US rollout before considering broader implementation.

‘This is just one part of our broader commitment to online safety for young users and families,’ the blog post reads. ‘We’ve continually invested in technology, policies, and literacy resources to better protect kids and teens across our platforms.’

Nonetheless, the programme is likely to attract scrutiny. Critics may question the accuracy of AI-powered age detection and whether the measures strike the right balance between safety, privacy, and personal autonomy — or risk overstepping.

Some parents and privacy advocates may also raise concerns about the level of visibility and control families will have over how children are identified and managed by the system.

As public pressure grows for tech firms to take greater responsibility in protecting vulnerable users, Google’s rollout may signal the beginning of a new industry standard.

The shift towards AI-based age assurance reflects a growing consensus that digital platforms must proactively mitigate risks for young users through smarter, more adaptive technologies.

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

VPN dangers highlighted as UK’s Online Safety Act comes into force

Britons are being urged to proceed with caution before turning to virtual private networks (VPNs) in response to the new age verification requirements set by the Online Safety Act.

The law, now in effect, aims to protect young users by restricting access to adult and sensitive content unless users verify their age.

Instead of offering anonymous access, some platforms now demand personal details such as full names, email addresses, and even bank information to confirm a user’s age.

Although the legislation targets adult websites, many people have reported being blocked from accessing less controversial content, including alcohol-related forums and parts of Wikipedia.

As a result, more users are considering VPNs to bypass these checks. However, cybersecurity experts warn that many VPNs can pose serious risks by exposing users to scams, data theft, and malware. Without proper research, users might install software that compromises their privacy rather than protecting it.

With Ofcom reporting that eight per cent of children aged 8 to 14 in the UK have accessed adult content online, the new rules are viewed as a necessary safeguard. Still, concerns remain about the balance between online safety and digital privacy for adult users.

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

White House launches AI Action Plan with Executive Orders on exports and regulation

The White House has unveiled a sweeping AI strategy through its new publication Winning the Race: America’s AI Action Plan.

Released alongside three Executive Orders, the plan outlines the federal government’s next phase in shaping AI policy, focusing on innovation, infrastructure, and global leadership.

The AI Action Plan centres on three key pillars: accelerating AI development, establishing national AI infrastructure, and promoting American AI standards globally. Four consistent themes run through each pillar: regulation and deregulation, investment, research and standardisation, and cybersecurity.

Notably, deregulation is central to the plan’s strategy, particularly in reducing barriers to AI growth and speeding up infrastructure approval for data centres and grid expansion.

Investment plays a dominant role. Federal funds will support AI job training, data access, lab automation, and domestic component manufacturing, instead of relying on foreign suppliers.

Alongside, the plan calls for new national standards, improved dataset quality, and stronger evaluation mechanisms for AI interpretability, control, and safety. A dedicated AI Workforce Research Hub is also proposed.

In parallel, three Executive Orders were issued. One bans ‘woke’ or ideologically biased AI tools in federal use, another fast-tracks data centre development using federal land and brownfield sites, and a third launches an AI exports programme to support full-stack US AI systems globally.

While these moves open new opportunities, they also raise questions around regulation, bias, and the future shape of AI development in the US.

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