Southwark Education, Learning and Achievement logo Southwark Education, Learning and Achievement
Sign in

Sign in to Southwark Education, Learning and Achievement or Complete a task any time of day with your dedicated, personalised account Create account

Google search and AI safety

A question was recently raised with the local authority about the safety of Google Search for children following the introduction of AI Overview and AI Mode. In response, LGfL has provided safeguarding and technical guidance for schools.

A recent assessment by Common Sense Media rated Google Search’s AI Overview and AI Mode as presenting an “Unacceptable Risk” for children and young people. The assessment found concerns relating to inaccurate information, inappropriate responses to safeguarding-related searches, unreliable source quality and the ability of AI Mode to complete pupils’ homework. It also noted that Google Search’s AI features cannot currently be switched off individually, although users can scroll past AI Overviews or choose not to enter AI Mode.

Google has challenged elements of the assessment, stating that it could not replicate many of the findings and that its AI Search features include safety protections. However, the report reinforces the need for schools to use clear safeguards, supervision and digital-literacy education when pupils use online search tools.

 

LGfL Advice

LGfL recommends that schools take a strategic, safeguarding-led approach to Generative AI. As AI is increasingly embedded within mainstream products, it cannot always be managed by simply blocking a single website or URL. Schools should consider the implications for pupils, staff, parents and leaders, including data protection, intellectual property, accuracy and online-safety risks.

LGfL’s AI Policy Toolkit provides practical support for developing a whole-school approach to AI. Updated template policies covering Generative AI risks are also available through LGfL Safe Policies.

 

Recommended Measures

LGfL advises schools to:

  • Adopt a block-by-default, allow-by-exception approach for AI tools, enabling access only where there is a clear educational purpose and appropriate safeguards.
  • Ensure user authentication is in place, allowing access to be managed appropriately by age group and user type.
  • Use HTTPS decryption where appropriate, as this supports more precise filtering of webpages and AI-enabled features.
  • Review Google Workspace settings to decide whether Gemini should be available within Google applications.
  • Apply monitoring arrangements proactively, including through tools such as Senso, to identify potential safeguarding concerns and understand how pupils are using online services.
  • Teach pupils to treat AI-generated answers critically, check sources and seek support from trusted adults where content raises concerns.

 

Filtering Review

LGfL also advises schools to review filtering arrangements regularly. In particular, the following categories should normally be blocked for pupils:

  • Host is an IP, as this can allow attempts to bypass filtering.
  • Games, unless specific educational sites or URLs have been approved.
  • Humour, which can include offensive or unsuitable material.
  • Sales, as general shopping sites may contain adult or inappropriate products.
  • Search Engine Other, where SafeSearch protections may not be available.
  • Social Media, unless access has been specifically assessed and approved.
  • Web Chat, due to potential exposure to adult content and contact risks.

This advice has been shared in response to the concerns raised and is intended to support schools in reviewing their current arrangements as AI features become more common in the tools pupils use every day.