---

Livestreamed child sexual abuse: Scale, detection and how it can be disrupted

Warning: This briefing discusses sexual abuse of children, and refers to specific actions that constitute criminal offences, which readers may find distressing.

DOI: https://doi.org/10.58248/PN776

Livestreaming technologies can allow people to share and view videos of child sexual abuse (CSA) with a perceived reduced risk of detection. Livestreams may be pre-recorded videos of abuse that are later livestreamed for an audience of offenders. Offenders may also entice children into using livestreaming platforms to self-generate child sexual abuse material (CSAM). They may also use livestreaming platforms to direct child abuse in real time.

The livestreaming of abuse often involves victims based in low-income countries, with offenders paying to view and direct the abuse from other parts of the world. This is also referred to as cybersex trafficking or live distant child sexual abuse (LDCA).

Livestreams may leave no stored file unless deliberately recorded, and abuse may take place through private or encrypted communication channels. Virtual private networks (VPNs) can also reduce visibility for law enforcement agencies and internet service providers. Technologies used to detect stored child sexual abuse material (CSAM) may therefore be less effective for livestreamed abuse.

Scale of offending

Evidence on the scale of livestreamed CSA is limited, but available information indicates significant UK demand. The Philippines is a known hotspot for LDCA and receives 3,000 reports of suspected cases of online child sexual exploitation from overseas every month. Analysis of suspicious financial transaction data from the Philippines identified the UK as a significant demand country. A recent analysis indicated that 1 in 71 men in the UK (or 1.4%) have ever “webcammed a child in a sexual way”.

Detection technologies and their limits

Platforms and internet service providers traditionally detect known CSAM using hash-matching, which compares potentially illegal files against databases of digital fingerprints, or "hashes", of confirmed CSAM. Automated artificial intelligence (AI) classifiers can also identify previously unseen CSAM.

However, these approaches are less effective where abuse is livestreamed, not stored, or takes place through private or encrypted one-to-one communications. AI-generated CSAM creates further challenges, including increased pressure on law enforcement agencies and difficulties in detection and classification.

Offenders can also sometimes be identified through financial data. Payments for LDCA may leave traceable records even where the livestream itself leaves little or no technical evidence. Public-private partnerships between law enforcement agencies, technology companies and financial institutions can support the identification and prosecution of UK offenders.

Regulation

The UK Online Safety Act 2023 covers user-to-user communications, which include social media, messaging applications, video games' messaging features, and livestreams. LDCA is typically broadcast on private or encrypted livestreams.

Under the act, Ofcom may recommend the use of a specific technology to analyse public user-generated content. However, it cannot do so for content communicated privately.

Ofcom has powers under section 121 of the act to require an online service provider to use a specific technology to tackle CSAM in private communications, where it decides it is necessary and proportionate. There are, however, a number of steps that would need to be taken under the act before exercising this power.

Other countries are considering or adopting different approaches to tackle livestreamed CSA, which may be relevant because LCSA often involves offenders, victims, facilitators, platforms and financial flows across national borders.

Acknowledgements

This briefing was produced in consultation with experts and stakeholders, who are listed at the end of the briefing. POST would like to thank everyone who contributed their expertise to this briefing.