The next UK Parliament can’t ignore AI and online safety

The King’s Speech on 12 May set out the UK Government’s legislative priorities for the next Parliament. There was little specific detail on online safety or AI regulation even though the risks posed by AI and digital platforms are only becoming more urgent.

Our research has repeatedly shown that online harms are evolving faster than safeguards to prevent them. Social media platforms continue to amplify abuse, extremism, and conspiracies through engagement-driven systems. At the same time, increasingly powerful AI chatbots are being deployed to millions of users without adequate protections in place.

The next Parliament has an opportunity to address these harms with practical, targeted reforms that build on existing legislation and ensure regulation keeps pace with technology.

The UK already has important foundations in place through the Online Safety Act. But emerging technologies and rapidly changing digital ecosystems mean further action is needed to close gaps, strengthen accountability, and ensure safety-by-design.

Join our email community to get regular updates on online safety regulation in the UK:

AI chatbot harms can no longer be treated as theoretical

In March this year, a teenager in the UK asked DeepSeek—a widely used AI chatbot—what weapon an inexperienced killer should use to murder his mother. He went on to follow DeepSeek’s advice and kill his mother.

This case is not an outlier. Our Killer Apps report shows that 8 in 10 popular chatbots are regularly willing to assist in violent planning — including providing advice on weapons, tactics and locations.

It’s not just mass violence. Around the world, we’re seeing more cases of self-harm and teen suicide. In Fake Friend, we found that ChatGPT advised a simulated teenage user on how to “safely” self-harm and generated a suicide note addressed to her parents.

These failures are not edge cases. They reflect systemic weaknesses in the safeguards surrounding products used daily by millions of people, including children and vulnerable users.

That is why CCDH joined survivors and bereaved families of terror attacks in calling on the next Parliament to act urgently on the extremism and national security risks posed by AI chatbots. In an open letter ahead of the King’s Speech, survivors warned that online algorithms increasingly fuel radicalisation and that AI systems are lowering the barriers to real-world violence by providing practical assistance to attackers.

Together, we urged the Government to introduce clear duties on AI chatbot developers to assess and mitigate extremism and terrorism risks, ensure transparency and independent oversight of high-risk systems, and strengthen implementation of the Online Safety Act.

These are not abstract policy debates. They are issues of public safety.

The Online Safety Act must evolve alongside technology

The Online Safety Act was created because technology companies cannot be relied upon to regulate themselves when commercial incentives reward engagement over safety.

Parliament must now build on the Act to ensure it remains effective against emerging harms.

One clear priority is closing loopholes around AI chatbots. Despite their growing influence over users’ behaviour, emotions, and decision-making, chatbots currently sit outside many of the safety duties already expected of social media platforms.

The next Parliament should ensure that chatbot providers are subject to meaningful risk assessment and mitigation obligations, particularly where foreseeable harms to children, vulnerable people, or public safety exist.

There is also an opportunity to strengthen the Act’s broader “safe-by-design” ambitions. Online platforms and AI systems should be expected to proactively identify and reduce foreseeable risks, rather than simply reacting once harms occur.

CCDH’s research consistently shows that many harms are not accidental. Platform design choices, recommendation systems, and engagement-based incentives actively shape how harmful content spreads online. Effective regulation must address those systemic risks directly.

Technology evolves rapidly. Regulation must evolve with it

In collaboration with partners across the Online Safety Act Network, CCDH and other campaigners have identified the growing range of online harms that require stronger oversight and made actionable policy proposals to address them.

Armed with this evidence, the next Parliament does not need to start from scratch. The UK already has legislative foundations through the Online Safety Act. What is needed now is the political will to close gaps in protection and ensure the law reflects how digital technologies actually function today.

That includes:

• Strengthening risk assessment and mitigation duties for AI chatbot developers
• Embedding meaningful safety-by-design obligations across digital services
• Ensuring platforms address systemic risks, not only isolated illegal content
• Improving transparency and independent oversight of high-risk AI systems
• Accelerating effective implementation and enforcement of the Online Safety Act

At CCDH, we will continue to document the failures of technology companies to protect users from foreseeable harm. But evidence alone is not enough.

As online harms continue to evolve, the next Parliament must ensure that public safety evolves with them.