Internet safety campaigners have urged Ofcom, the UK’s communications regulator, to limit Meta’s use of AI in risk assessments.

This follows reports that up to 90% of these could soon rely on AI.

The UK’s Online Safety Act requires platforms to assess harms, especially around child safety and illegal content.

Groups like the Molly Rose Foundation, NSPCC, and Internet Watch Foundation raised concerns in a letter to Ofcom chief Melanie Dawes, warning that AI-led assessments could compromise the act’s standards.

The letter asked Ofcom to challenge the idea that platforms can reduce risk assessment processes. It urged that AI-led assessments should not be seen as adequate.

Ofcom said it is reviewing the concerns and will respond in due course.

Meta denied using AI to make direct risk decisions, stating that AI tools assist human teams in meeting legal and policy needs.

🔗 Source: The Guardian

🧠 Food for thought

1️⃣ Industry-wide tension between AI efficiency and human oversight in risk management

The Meta controversy reflects a broader pattern across industries where organizations increasingly turn to AI for risk assessment while facing questions about appropriate human oversight.

Research shows AI-driven risk tools claim significant efficiency gains, with some solutions reducing administrative workload by 50-70% and workplace incidents by up to 35%1.

However, industry frameworks like NIST’s AI Risk Management Framework emphasize the importance of human involvement in critical functions: Identify, Assess, Manage, and Communicate2.

This tension is particularly acute for social media platforms where complex harm assessments involving children require nuanced judgment that purely automated systems may struggle to provide.

Organizations adopting AI for risk management must balance efficiency gains against the need for human expertise, especially when legal compliance and user safety are at stake.

2️⃣ UK’s AI regulatory approach is evolving from principles to enforcement

Recent Meta developments