Tech brief
Regulatory Clashes and Agentic Ambitions
UK pressure mounts on TikTok's child safety as DeepMind targets enterprise AI costs and OpenAI's supercomputer lobbying draws fire.
Tech
Ofcom's formal probe of TikTok is the Online Safety Act's first real test — forcing platforms to fix verification gaps or face massive fines before next year's social media ban.
BackgroundThe Online Safety Act 2023 requires digital platforms to use highly effective age assurance systems to prevent children from encountering harmful material online. Failure to comply can result in severe financial penalties and direct regulatory intervention.
- Ofcom's investigation follows a statutory report warning that substantial verification gaps remain despite platforms increasing their age checks in late 2025, suggesting voluntary measures are falling short.
- The findings are particularly sensitive as the UK government plans to enforce a total social media ban for under-16s starting next year, elevating TikTok's compliance from a regulatory hurdle to an existential threat.
- TikTok faces potential fines of up to 10% of its global revenue if the regulator finds it systematically breached its safety duties, which could trigger a massive legal battle over algorithmic safety obligations.
Tech
DeepMind's lightweight models slash AI agent token costs by 65% — clearing the primary financial barrier to widespread enterprise automation.
BackgroundProduction AI agents require frequent API calls to perform complex, long-horizon tasks like coding or vulnerability patching. High token costs and slow response times have historically acted as major barriers to enterprise deployment of these autonomous systems.
- Gemini 3.6 Flash halves the processing time per task and reduces token usage by up to 65% in long-horizon engineering operations, making autonomous agents economically viable.
- Gemini 3.5 Flash-Lite is designed for high-throughput tasks, generating 350 output tokens per second to enable instantaneous customer service responses.
- The specialized Gemini 3.5 Flash Cyber model is integrated into Google's CodeMender agent, helping developers and IT teams autonomously patch software vulnerabilities before they are exploited.
Tech
OpenAI's public attacks on British planning regulations have sparked political backlash — exposing its paused £30 billion supercomputer investment as a speculative lobbying leverage play.
BackgroundLarge-scale artificial intelligence models require massive data centers that consume vast amounts of electricity and water. These projects often face local environmental opposition and strict grid capacity constraints in highly populated areas like the UK.
- Activists expressed anger at the OpenAI employee's public comments at the right-wing political event, claiming the firm is pressuring the UK to loosen environmental protections.
- A recent Freedom of Information investigation revealed that OpenAI never actually visited key proposed sites for the heavily promoted £30 billion investment, raising doubts about its commitment.
- Former ministers had widely publicized the potential Stargate UK project, but documents suggest that £20 billion of the proposed funding was entirely hypothetical, undermining the government's digital infrastructure claims.
Unlock the full brief
Sign in to read every signal, takeaway, and source. Free account — Apple, Google, or email.