Invisible Predators: What South African School Heads Need to Know About AI and Safeguarding

A South African school can complete every expected safeguarding check and still leave material risk unseen. Police clearance certificates, references and interviews remain essential, but they have limits–particularly when harmful behaviour has not been reported or connected to a person’s professional identity. Digital anonymity and AI are widening these gaps.

TL;DR

  • A completed safeguarding process can still miss material risk.
  • Digital anonymity and AI allow harmful behaviour to sit alongside a credible professional identity.
  • Traditional checks remain important, but each provides only a partial view.
  • South African school leaders need layered assurance, modern governance and human judgement.

Professional credibility does not remove hidden risk

In 2022, Matthew Smith was appointed deputy head of pastoral care and safeguarding at Thomas’s School in London.

His checks were clean, his references were glowing and he had worked in respected schools and orphanages. He appeared to be exactly the kind of person a school could trust.

He was later arrested for paying for child sexual abuse to be committed and filmed to order. In 2023, he was sentenced to 12 years in prison.

Although this happened in the UK, the warning applies directly to South African schools: Smith passed the expected checks, yet material risk remained unseen.

Compliance confirms that the required process took place. It does not necessarily confirm that the process provides the assurance placed upon it.

The hidden internet has changed the risk environment

Beyond the surface web - indexed by search engines - is a much larger body of unindexed content. A small part is the dark web, which requires specialist software such as Tor.

This technology has legitimate uses, but its anonymity can also conceal identity, communications and illegal activity. Someone operating in this hidden environment can simultaneously maintain a completely normal professional identity in the visible world.

How AI increases the risk

AI has reduced the cost, skill and time required to deceive. It can create convincing false documents and identities, synthetic images, imitated voices and personalised messages at scale.

A false police clearance certificate can be generated and visually aged in minutes. A qualification can be fabricated, and a reference conversation may not involve the person a school believes it is speaking to.

Completing a check and verifying its underlying evidence are no longer necessarily the same thing.

What recent South African cases show

The documentary School Ties examined grooming and sexual abuse by trusted adults at South African schools.

In November 2025, Parliament’s Portfolio Committee on Basic Education heard that 469 cases of educator sexual misconduct had been recorded between 2018/19 and 2024/25. It called for stronger coordination and a centralised national database because individuals can leave one institution and resurface elsewhere. Read the committee’s statement.

Statutory checks remain essential, but a clean result should not be mistaken for complete assurance.

The modern vetting blind spot

Every traditional layer is useful. The risk is expecting more assurance than those layers can provide.

Police clearance certificates

A police clearance tells you what is known to the authorities at that point in time. It cannot reveal behaviour that has not been detected, reported or linked to that person. It also provides limited assurance if the document itself has not been independently verified.

CVs, qualifications and references

These tell you what a candidate claims and what another organisation will confirm. Documents can be fabricated, reference processes manipulated and relevant concerns dispersed across disconnected records.

Interviews

An interview shows how someone chooses to present. Confidence, warmth and a convincing professional story are not independent evidence.

Google searches

A Google search can reveal useful indexed information, but not the full online environment. Someone can curate a clean surface-web presence while operating elsewhere under another identity.

All these layers remain important. The mistake is assuming that completing every step must reveal the whole story.

Five practical actions for South African heads and governors

1. Challenge your assumptions

Ask not only whether a check was completed, but what it proves, how the evidence was verified and what remains outside its scope.

2. Build layered assurance

Police clearances, identity and qualification verification, references, interviews, safeguarding culture and appropriate digital intelligence should complement one another.

The more sensitive the role and the greater the access to learners, the stronger the assurance should be.

3. Modernise governance

AI and digital risk cannot sit only with IT or the safeguarding lead. Heads, Excos and governing bodies should reflect emerging threats in policy, recruitment, training and risk oversight.

4. Create a culture that surfaces weak signals

Serious cases rarely begin with one dramatic warning. Strong safeguarding cultures allow early concerns to be raised–even about people who are popular, successful or senior.

Trust must never become immunity from scrutiny.

5. Prepare before the incident

Schools should decide in advance who leads the response, involves authorities, communicates with families and supports the learner.

Questions for your Exco or governing body

Use these questions to test whether your process still provides the assurance you expect:

  • Does our safeguarding process give us the assurance we think it does?
  • Does scrutiny match each role’s level of trust and access to learners?
  • How would we verify a police clearance certificate, qualification or reference?
  • Can staff raise concerns about a popular or senior colleague?
  • Who does what during the first 24 hours of an incident?

South African heads and governing bodies should own these questions.

Leadership in an age of abundant information

AI is changing education, opportunity and risk. It can improve coverage and challenge assumptions, but leaders still have to decide what is relevant, fair and proportionate.

People remain responsible for safeguarding decisions. School governance and assurance should reflect the environment as it now exists, rather than the one in which the original process was designed.

An additional layer of visibility

Safehire.ai provides Digital Risk Screening for South African schools, using AI-led searches across surface, deep and dark-web sources with findings validated by human analysts.

It adds visibility alongside statutory checks; it does not replace them. Judgement remains with people.

Take one practical question to your next Exco or governing-body meeting:

Does our current process give us the assurance we think it does?

Continue reading