Interdisciplinary approaches Archives | Page 60 of 194

AI bot swarms emerge as a new threat to democracy

Researchers and free-speech advocates are warning that coordinated swarms of AI agents could soon be deployed to manipulate public opinion at a scale capable of undermining democratic systems.

According to a consortium of academics from leading universities, advances in generative and agentic AI now enable large numbers of human-like bots to infiltrate online communities and autonomously simulate organic political discourse.

Unlike earlier forms of automated misinformation, AI swarms are designed to adapt to social dynamics, learn community norms and exchange information in pursuit of a shared objective.

By mimicking human behaviour and spreading tailored narratives gradually, such systems could fabricate consensus, amplify doubt around electoral processes and normalise anti-democratic outcomes without triggering immediate detection.

Evidence of early influence operations has already emerged in recent elections across Asia, where AI-driven accounts have engaged users with large volumes of unverifiable information rather than overt propaganda.

Researchers warn that information overload, strategic neutrality and algorithmic amplification may prove more effective than traditional disinformation campaigns.

The authors argue that democratic resilience now depends on global coordination, combining technical safeguards such as watermarking and detection tools with stronger governance of political AI use.

Without collective action, they caution that AI-enabled manipulation risks outpacing existing regulatory and institutional defences.

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

Grok faces regulatory scrutiny in South Korea over explicit AI content

South Korea has moved towards regulatory action against Grok, the generative AI chatbot developed by xAI, following allegations that the system was used to generate and distribute sexually exploitative deepfake images.

The country’s Personal Information Protection Commission has launched a preliminary fact-finding review to assess whether violations occurred and whether the matter falls within its legal remit.

The review follows international reports accusing Grok of facilitating the creation of explicit and non-consensual images of real individuals, including minors.

Under the Personal Information Protection Act of South Korea, generating or altering sexual images of identifiable people without consent may constitute unlawful handling of personal data, exposing providers to enforcement action.

Concerns have intensified after civil society groups estimated that millions of explicit images were produced through Grok over a short period, with thousands involving children.

Several governments, including those in the US, Europe and Canada, have opened inquiries, while parts of Southeast Asia have opted to block access to the service altogether.

In response, xAI has introduced technical restrictions preventing users from generating or editing images of real people. Korean regulators have also demanded stronger youth protection measures from X, warning that failure to address criminal content involving minors could result in administrative penalties.

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

France accelerates rapid ban on social media for under-15s

French President Emmanuel Macron has called for an accelerated legislative process to introduce a nationwide ban on social media for children under 15 by September.

Speaking in a televised address, Macron said the proposal would move rapidly through parliament so that explicit rules are in place before the new school year begins.

Macron framed the initiative as a matter of child protection and digital sovereignty, arguing that foreign platforms or algorithmic incentives should not shape young people’s cognitive and emotional development.

He linked excessive social media use to manipulation, commercial exploitation and growing psychological harm among teenagers.

Data from France’s health watchdog show that almost half of teenagers spend between two and five hours a day on their smartphones, with the vast majority accessing social networks daily.

Regulators have associated such patterns with reduced self-esteem and increased exposure to content linked to self-harm, drug use and suicide, prompting legal action by families against major platforms.

The proposal from France follows similar debates in the UK and Australia, where age-based access restrictions have already been introduced.

The French government argues that decisive national action is necessary instead of waiting for a slower Europe-wide consensus, although Macron has reiterated support for a broader EU approach.

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

New phishing attacks exploit visual URL tricks to impersonate major brands

Generative phishing techniques are becoming harder to detect as attackers use subtle visual tricks in web addresses to impersonate trusted brands. A new campaign reported by Cybersecurity News shows how simple character swaps create fake websites that closely resemble real ones on mobile browsers.

The phishing attacks rely on a homoglyph technique where the letters ‘r’ and ‘n’ are placed together to mimic the appearance of an ‘m’ in a domain name. On smaller screens, the difference is difficult to spot, allowing phishing pages to appear almost identical to real Microsoft or Marriott login sites.

Cybersecurity researchers observed domains such as rnicrosoft.com being used to send fake security alerts and invoice notifications designed to lure victims into entering credentials. Once compromised, accounts can be hijacked for financial fraud, data theft, or wider access to corporate systems.

Experts warn that mobile browsing increases the risk, as users are less likely to inspect complete URLs before logging in. Directly accessing official apps or typing website addresses manually remains the safest way to avoid falling into these traps.

Security specialists also continue to recommend passkeys, strong, unique passwords, and multi-factor authentication across all major accounts, as well as heightened awareness of domains that visually resemble familiar brands through character substitution.

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

Generative AI targets antibiotic resistance with rapid drug design

At the World Economic Forum, scientists warned that deaths from drug-resistant ‘superbugs,’ microbes that can withstand existing antibiotics, may soon exceed fatalities from cancer unless new treatments are found.

To address this, companies like Basecamp Research have developed AI models trained on extensive genetic and biological data to accelerate drug discovery for complex diseases, including antibiotic resistance.

These AI systems can design novel molecules predicted to be effective against resistant microbes, with early laboratory testing showing a high success rate for candidates suggested by the models.

The technology enables a user to prompt the system to design entirely new molecular structures that bacteria have never encountered, potentially yielding treatments capable of combating resistant strains.

The approach reflects a broader trend in using AI for biomedical discovery, where generative models reduce the time and cost of identifying new drug candidates. While still early and requiring further validation, such systems could reshape how antibiotics are developed, offering new tools in the fight against antimicrobial resistance.

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

Writers challenge troubling AI assumptions about language and style

A growing unease among writers is emerging as AI tools reshape how language is produced and perceived. Long-established habits, including the use of em dashes and semicolons, are increasingly being viewed with suspicion as machine-generated text becomes more common.

The concern is not opposition to AI itself, but the blurring of boundaries between human expression and automated output. Writers whose work was used to train large language models without consent say stylistic traits developed over decades are now being misread as algorithmic authorship.

Academic and editorial norms are also shifting under this pressure. Teaching practices that once valued rhythm, voice, and individual cadence are increasingly challenged by stricter stylistic rules, sometimes framed as safeguards against sloppy or machine-like writing rather than as matters of taste or craft.

At the same time, productivity tools embedded into mainstream software continue to intervene in the writing process, offering substitutions and revisions that prioritise clarity and efficiency over nuance. Such interventions risk flattening language and discouraging the idiosyncrasies that define human authorship.

As AI becomes embedded in publishing, education, and professional writing, the debate is shifting from detection to preservation. Many writers warn that protecting human voice and stylistic diversity is essential, arguing that affectless, uniform prose would erode creativity and trust.

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

LinkedIn phishing campaign exposes dangerous DLL sideloading attack

A multi-faceted phishing campaign is abusing LinkedIn private messages to deliver weaponised malware using DLL sideloading, security researchers have warned. The activity relies on PDFs and archive files that appear trustworthy to bypass conventional security controls.

Attackers contact targets on LinkedIn and send self-extracting archives disguised as legitimate documents. When opened, a malicious DLL is sideloaded into a trusted PDF reader, triggering memory-resident malware that establishes encrypted command-and-control channels.

Using LinkedIn messages increases engagement by exploiting professional trust and bypassing email-focused defences. DLL sideloading allows malicious code to run inside legitimate applications, complicating detection.

The campaign enables credential theft, data exfiltration and lateral movement through in-memory backdoors. Encrypted command-and-control traffic makes containment more difficult.

Organisations using common PDF software or Python tooling face elevated risk. Defenders are advised to strengthen social media phishing awareness, monitor DLL loading behaviour and rotate credentials where compromise is suspected.

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

Stanford and Swiss institutes unite on open AI models

Stanford University, ETH Zurich, and EPFL have launched a transatlantic partnership to develop open-source AI models prioritising societal values over commercial interests.

The partnership was formalised through a memorandum of understanding signed during the World Economic Forum meeting in Davos.

The agreement establishes long-term cooperation in AI research, education, and innovation, with a focus on large-scale multimodal models. The initiative aims to strengthen academia’s influence over global AI by promoting transparency, accountability, and inclusive access.

Joint projects will develop open datasets, evaluation benchmarks, and responsible deployment frameworks, alongside researcher exchanges and workshops. The effort aims to embed human-centred principles into technical progress while supporting interdisciplinary discovery.

Academic leaders said the alliance reinforces open science and cultural diversity amid growing corporate influence over foundation models. The collaboration positions universities as central drivers of ethical, trustworthy, and socially grounded AI development.

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

Google adds Personal Intelligence to AI Search

Google has expanded AI Search with Personal Intelligence, enabling more personalised responses using Gmail and Google Photos data. The feature aims to combine global information with individual context to deliver search results tailored to each user.

Eligible Google AI Pro and AI Ultra subscribers can opt in to securely connect their Gmail and Photos accounts, allowing Search to draw on personal preferences, travel plans, purchases, and memories.

The system uses contextual insights to generate recommendations that reflect users’ habits, interests, and upcoming activities.

Personal Intelligence enhances shopping, travel planning, and lifestyle discovery by anticipating needs and offering customised suggestions. Privacy controls remain central, with users able to manage data connections and turn off personal context at any time.

The feature is launching as an experimental Labs release for English-language users in the United States, with broader availability expected following testing. Google said ongoing feedback will guide refinements as the system continues to evolve.

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

Education for Countries programme signals OpenAI push into public education policy

OpenAI has launched the Education for Countries programme, a new global initiative designed to support governments in modernising education systems and preparing workforces for an AI-driven economy.

The programme responds to a widening gap between rapid advances in AI capabilities and people’s ability to use them effectively in everyday learning and work.

Education systems are positioned at the centre of closing that gap, as research suggests a significant share of core workplace skills will change by the end of the decade.

By integrating AI tools, training and research into schools and universities, national education frameworks can evolve alongside technological change and better equip students for future labour markets.

The programme combines access to tools such as ChatGPT Edu and advanced language models with large-scale research on learning outcomes, tailored national training schemes and internationally recognised certifications.

A global network of governments, universities and education leaders will also share best practices and shape responsible approaches to AI use in classrooms.

Initial partners include Estonia, Greece, Italy, Jordan, Kazakhstan, Slovakia, Trinidad and Tobago and the United Arab Emirates. Early national rollouts, particularly in Estonia, already involve tens of thousands of students and educators, with further countries expected to join later in 2026.

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