AI Regulation & Policy

Today's Stories

1. 1

   Terrorist groups using ChatGPT, Claude, Gemini to plan attacks and build weapons

   ISIS has trained Boko Haram commanders to bypass AI safety filters since 2023. Both Boko Haram factions now use popular AI chatbots including ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek, and have set up dedicated AI units. The group uses AI for attack planning, building explosives, weapons maintenance, and operational security, according to a study by researcher Antonia Jülich of the Cambridge Programme on AI Science & Policy based on 57 interviews with 27 former members. Safety filters on major chatbots failed to reliably prevent misuse, even though AI researchers and companies like OpenAI and Anthropic have long warned about this risk. Jülich notes that while Boko Haram's current AI use remains conventional, former members described enthusiasm for the technology and some said the group had previously considered mass-casualty weapons—a warning that terrorists may pursue AI assistance for chemical and biological weapons.

   Researchers point out that widely-used chatbots like ChatGPT and Claude mostly make existing knowledge easier to access rather than generating anything new. The greater concern is potential misuse of more specialized AI systems in the life sciences.

2. 2

   Taiwan's AI shift to production signals global enterprise adoption turning point

   Taiwan is moving its AI deployments from experimental trials to production systems at scale, signaling a broader shift in how companies worldwide are adopting automation and AI agents. As firms deploy AI agents into real operations, the challenge is no longer whether to experiment with AI—it is how to ensure control, reliability, and measurable returns. This transition reflects a maturation in enterprise AI use that could reshape how data governance and digital security are managed across industries.

   The shift underscores that success in AI adoption now depends on moving beyond testing and into operational systems that deliver measurable business value.

3. 3

   DeepSeek's open-weights model ranks #2 in reasoning, reshaping AI competition

   DeepSeek released an open-weights reasoning model that ranks #2 among open-weights reasoning models, according to benchmarks cited in the article. The model operates on 27% of the FLOPs (computational power) compared with DeepSeek-V3.2 and uses 83.9 GiB, positioning it as a notably efficient alternative in the reasoning model category. The #2 ranking and lower computational requirements signal that high-performance AI reasoning is becoming accessible beyond proprietary systems, potentially reshaping how businesses and developers choose their AI infrastructure. For companies evaluating reasoning models, this development suggests competitive options exist at different efficiency and cost points.

   The model's efficiency (27% of FLOPs vs. DeepSeek-V3.2's baseline) and specific memory footprint (83.9 GiB) are the defining technical metrics; broader adoption patterns and how other model providers respond will indicate whether open-weights reasoning models become standard in enterprise AI tooling.

4. 4

   UN AI Summit Grapples With Tech Sector's Uneven Promises

   The United Nations' International Telecommunication Union (ITU) held its 10th annual AI for Good summit in Geneva, bringing together private and public sector representatives to discuss how artificial intelligence could address global challenges like hunger, disease, and climate change. The conference featured debates on access to AI models and computing resources, and announced formation of a 44-member commission cochaired by Rwandan president Paul Kagame and Salesforce CEO Marc Benioff to shepherd AI for Good initiatives. Participants raised serious concerns about how AI is being deployed without sufficient oversight, with humanitarian organizations warning that unchecked corporate control is already hardwiring global inequality and eroding human rights. Access to AI compute and models remains unevenly distributed—most large language models are structured around English, and export controls and intellectual property restrictions risk excluding poorer countries from shaping the technology's future. Engineers and standards bodies are now recognizing that human rights and equity are core infrastructure questions, not afterthoughts.

   The divide between rhetorical commitment and concrete action persisted throughout the summit. While humanoid robots and Tesla Cybertrucks were displayed on the convention floor, attendees and speakers emphasized that technical decisions—built into hidden architecture, technical standards, and procurement choices—matter far more than policy declarations. The challenge now is translating high-level human rights principles into verifiable technical enforcement and practical impact assessments with real accountability, rather than governance theater.