Chapter 19
Scams, fakes and everyday risks
The chapter in a minute
- Deepfakes and copied voices of real people are already used to steal money. When a face or voice you know asks for money in a hurry, hang up and ring back on a number you already have. A family safe word is a second check.
- A scam message may have no spelling mistakes now. Judge it by what it wants from you: a link, a payment or your details.
- Leaks come from faults and from sharing, and accounts and agents are worth tricking. Leave private details out, treat a shared chat link as public, give an agent only the access the job needs and never give anyone your password.
- Before a child uses a chat assistant, look up two things: its minimum age, and whether it needs your permission.
- Report intimate images shared without consent, and never forward, save or screenshot a sexual image of a child.
Try it now
Set a family safe word
0 of 6 steps done
Step 1 of 6
Make a short list of the family or friends who might ever ring you for help, and the ones you might ring.
Try this
Try this, from the book
Cover the numbered notes on Figure 19.2 with your hand and read the message on its own. Find three signs that it's a scam, then uncover the notes and compare. Next time a text asks you to pay or sign in, give it the same three-sign check before you tap anything!
Tools and settings, as of a date
The book points to the ‘Scams and safety’ section of the companion page. Here it is in full, with the date each detail was checked. Open it on the companion page
Chapter 19 points here. Every detail below was read on its source page on 2 November 2026 unless another date is given.
Dated case list
- Hong Kong, early 2024 (reported 26 June 2024). Police in Hong Kong reported a case in which a pre-recorded video conference was faked using downloaded online public video clips and voices. An employee authorised payments to five local accounts, a loss of about HK$200 million. Hong Kong government press release (opens in a new tab)
- Northern Ireland, September 2026. The Police Service of Northern Ireland warned about a fraud in which the victim lost £250,000 after seeing a video that appeared to be made with AI, of a well-known figure from the financial sector advertising an investment. Police say an advert or apparent celebrity endorsement "should never be regarded as proof" that an investment is genuine, and "Never allow yourself to be rushed into making a financial decision." PSNI news release, 7 September 2026 (opens in a new tab)
The book keeps the hedge "appeared to be" and names no individual and no company.
Ireland: where to report and get advice
- An Garda Síochána, "Fraud (opens in a new tab)".
- FraudSMART, run by Banking and Payments Federation Ireland with its member banks, "Phone Scams (opens in a new tab)". FraudSMART is an advice site, not a place to report a scam. Its phone scams page says that if an unexpected caller asks for remote access or personal details, you should "hang up immediately and report to your bank or service provider."
- Competition and Consumer Protection Commission, "How to spot a scam (opens in a new tab)", 23 April 2026. It warns that if you paid by bank transfer, the money is unlikely to come back. Ring your bank and report it all the same.
The main numbers are in "Where to get help" below.
UK: where to report
As of 2 November 2026, read on each body's own site:
- Report Fraud, for England, Wales and Northern Ireland: www.reportfraud.police.uk (opens in a new tab) or 0300 123 2040. Its contact page (opens in a new tab) says the phone line is open 24 hours a day, 7 days a week, with fewer staff between 8pm and 8am, and calls cost your standard network rate. There is a textphone on 0300 123 2050 and an automated chatbot at any hour. (The same page gives web chat hours as both "8am to 8pm" and "24 hours a day", so check the chat button on the day.) The service is run by the City of London Police. It replaced Action Fraud on 4 December 2025, according to the Gwent Police and Crime Commissioner's notice (opens in a new tab), with a full public launch in January 2026.
- Scotland: Report Fraud's FAQs (opens in a new tab) say "Police Scotland have not signed up to the Report Fraud process", so in Scotland call Police Scotland on 101. Police Scotland (opens in a new tab)
- 159 reaches your own bank or payment provider. Stop! Think Fraud (GOV.UK) (opens in a new tab)
- Scam texts: forward them free to 7726, your mobile provider's reporting number. You can also report a suspicious call to your mobile by texting the word "Call" and the number to 7726.
- Scam emails: forward them to report@phishing.gov.uk (opens in a new tab), the Suspicious Email Reporting Service. Both routes are on Report Fraud's page "How to report suspicious messages, calls and websites (opens in a new tab)" and on Stop! Think Fraud (opens in a new tab).
Shared links and chat history
As of 2 November 2026, from each provider's own help pages. Menu names change, so if a step doesn't match, look for the same words nearby.
ChatGPT (OpenAI). Shared link pages "are not intended for search-engine indexing, but this does not make a link private. Anyone who can access a link can also forward it." (OpenAI, sharing conversations (opens in a new tab), page updated 22 October 2026)
To see and delete your shared links:
- Go to Settings.
- Select Data controls.
- Next to Shared links, select Manage.
- Select Delete shared link next to one item, or open More actions and select Delete all shared links.
To delete a chat (OpenAI, deleting and archiving chats (opens in a new tab)):
- Open your chat history.
- Select the more options menu (•••) beside the chat.
- Select Delete, then confirm.
Archiving is not deleting. To delete everything, go to Settings, then Data controls, then Delete all chats; this can't be undone. A deleted chat leaves your account straight away and is scheduled for permanent deletion within 30 days, with security and legal exceptions.
Claude (Anthropic). Chats are private until you share one. A public link shows a snapshot of the chat as it was when you shared it, and anyone with the link can open it without an account. Each shared page asks search engines not to list it, but the company says "treat a public link as public". (Anthropic, public links (opens in a new tab))
To see and turn off your shared chats (Anthropic, share and unshare chats (opens in a new tab)):
- Go to Settings, then Privacy.
- Next to Shared chats, click Manage.
- Click Unshare next to any chat to stop its link working.
You can also open the chat's Share menu and change it from Public to Private.
To delete a chat on the web (Anthropic, delete or rename a conversation (opens in a new tab)):
- Find the chat in the left sidebar, or click "Chats and tasks".
- Hover over it and click the "⋮" button.
- Select Delete, then click Delete again to confirm.
To delete several, click "Chats and tasks", open the "⋮" menu on one chat, click Select, tick the others (or Select all), then Delete. A deleted chat leaves your history straight away and Anthropic's back-end systems within 30 days.
Gemini (Google). Anyone with a public link can read the chat and pass it on. (Google, share your chats (opens in a new tab))
To see and delete your public links:
- On a computer, go to gemini.google.com.
- At the bottom, click Settings & help, then Your public links.
- Click Delete public link at the end of a link, or Delete all links.
To delete your chat history (Google, manage and delete your Gemini Apps activity (opens in a new tab)):
- Go to gemini.google.com and click the Activity icon.
- Click Delete and choose All time, Last hour, Last day or Custom range; or click the delete icon next to a single chat.
- To set automatic deletion, click "Choose an auto-delete option" and pick 3, 18 (the default) or 36 months, or don't auto-delete.
Even with activity saving off, Google keeps your conversations for up to 72 hours to run the service.
Microsoft Copilot at work. Your organisation controls a work account, so ask your IT team first. To delete your Copilot activity history yourself (Microsoft, delete your Copilot activity history (opens in a new tab)):
- Sign in to the My Account portal with your work or school account.
- Go to Data and Privacy, then Data options, and open Copilot activity history.
- Select Delete, tick the Microsoft Copilot box, then select Delete again.
Microsoft says full deletion "might take some time", and files you saved are not removed.
Minimum ages, parental permission and the UK Children's code
See the table under "Prompt library: learning" (checked 23 October 2026), and the two legal consent ages there. Also: the UK Information Commissioner's Office expects services children are likely to use to be built with children in mind and to start with privacy settings high. "Age appropriate design: a code of practice for online services (opens in a new tab)". (The ICO's site put up a bot check on 2 November 2026, so this page was last read in full on 23 October 2026.)
Detection study figures
The book prints none of these; tools improve and figures date.
- Voices. Mai and colleagues, "Warning: Humans cannot reliably detect speech deepfakes (opens in a new tab)", PLOS ONE, 2 August 2023. 529 people listened to real and fake speech in English and Mandarin and "only correctly spotted the deepfakes 73% of the time". Hearing examples first helped only a little.
- All kinds of fakes. Diel and colleagues, "Human performance in detecting deepfakes: A systematic review and meta-analysis of 56 papers", Computers in Human Behavior Reports, volume 16, December 2024 (DOI 10.1016/j.chbr.2024.100538 (opens in a new tab); abstract read on the lead author's site (opens in a new tab)). Across 56 papers and 86,155 participants, people spotted fakes 55.54% of the time overall, not reliably better than chance: audio 62%, video 57%, images 53% and text 52%. Training, feedback and AI help raised this to about 65%.
- Detectors. Chandra and colleagues, "Deepfake-Eval-2024: A Multi-Modal In-the-Wild Benchmark of Deepfakes Circulated in 2024 (opens in a new tab)", first version 4 March 2025, current version 27 May 2026. On real fakes gathered from 88 websites in 52 languages, free detection tools scored about half as well as on older tests: their accuracy measure (AUC) fell by 50% for video, 48% for audio and 45% for images. Paid tools did better but "do not yet reach the accuracy of deepfake forensic analysts".
Checked: 2 November 2026.
Irish law on intimate images, and the open question
- Harassment, Harmful Communications and Related Offences Act 2020 (opens in a new tab) (sections 1 to 4 and 13, text as enacted; the whole Act commenced 10 February 2021).
- What the Minister told the Dáil. On 10 February 2026 the Minister for Justice, Home Affairs and Migration said that at present it is an offence to generate such images "provided they are then distributed and-or published", and that the Attorney General had confirmed the Act's definition of an intimate image covers an image of a person made with AI and altered images (deepfakes). He also said a five-year sentence for making an image "that is never spoken about and is never shared with anyone" would be "highly unusual", and the Government did not oppose the Bill at Second Stage. Dáil debates, 10 February 2026, pages 552 to 554 of the printed record (opens in a new tab). This official record replaces the newspaper report used in the notes (Irish Times, 16 January 2026 (opens in a new tab)).
- Bills before the Oireachtas, as of 2 November 2026. None has become law:
- Bill 8 of 2026 (opens in a new tab), a Private Members' Bill to "prohibit the creation of non-consensual intimate or harmful imagery": past Dáil Second Stage, no later stage listed (page last updated 10 February 2026).
- Bill 13 of 2026 (opens in a new tab), a Private Members' Bill to extend the law to images "generated or modified by computer-graphics": at Dáil Second Stage since 4 February 2026.
- Bill 96 of 2026 (opens in a new tab), a Private Members' Bill in the Seanad, amending the same Act and the Broadcasting Act 2009: at Seanad Second Stage since 13 October 2026.
- EU AI Act. In the consolidated text of 27 July 2026 (opens in a new tab) (still the latest consolidated version on EUR-Lex on 2 November 2026; amended by Regulation (EU) 2026/1744 of 8 July 2026), Article 5(1)(ba) bans systems that make realistic intimate images or videos of an identifiable person without their explicit consent, and Article 113 applies that ban from 2 December 2026. The labelling duty applies from 2 August 2026 (2 December 2026 for older systems, Article 111(4)), and Article 2(10) says the Act doesn't apply to people using AI in a "purely personal non-professional activity".
Checked: 2 November 2026.
Where to report intimate image abuse and illegal images of children
The chapter's routes, each confirmed on the body's own site on 2 November 2026:
- Ireland: Hotline.ie (opens in a new tab) or An Garda Síochána at your local station.
- UK: the police. For a sexual image of a child, the Internet Watch Foundation (opens in a new tab) or CEOP (opens in a new tab). Adults whose intimate images have been shared can get help from the Revenge Porn Helpline (opens in a new tab).
Never forward, save or screenshot a sexual image of a child, even as evidence. Note where you saw it and report it. Full details are under "Where to get help".
From the book
If a call, voice message or video asks for money in a hurry, hang up and ring back on a number you already have. Use the one saved in your phone, written in your address book or printed on your bank card. Never use a number the caller gives you.
Sources
- Consumer Terms of Service (opens in a new tab) · Anthropic, 2025
- Careful adoption of agentic AI services (opens in a new tab) · Australian Signals Directorate's Australian Cyber Security Centre and Cybersecurity and Infrastructure Security Agency and National Security Agency and Canadian Centre for Cyber Security and National Cyber Security Centre (New Zealand) and National Cyber Security Centre (UK), 2026
- Phone Scams (opens in a new tab) · Banking and Payments Federation Ireland, 2026
- C2PA and Content Credentials Explainer (opens in a new tab) · Coalition for Content Provenance and Authenticity, 2026
- How to spot a scam (opens in a new tab) · Competition and Consumer Protection Commission, 2026
- CEOP Safety Centre (opens in a new tab) · National Crime Agency, 2026
- Deepfake-Eval-2024: A Multi-Modal In-the-Wild Benchmark of Deepfakes Circulated in 2024 (opens in a new tab) · Chandra, Nuria Alina and Lee, Hannah and Murtfeldt, Ryan and Qiu, Lin and Karmakar, Arnab and Tanumihardja, Emmanuel and Farhat, Kevin and Caffee, Ben and Lee, Changyeon and Choi, Jongwook and Paik, Sejin and Kim, Aerin and Etzioni, Oren, 2025
- 'Concerns remain' around Grok despite X's assurances, says Minister (opens in a new tab) · Coyne, Ellen, 2026
- Human performance in detecting deepfakes: A systematic review and meta-analysis of 56 papers (opens in a new tab) · Diel, Alexander and Lalgi, Tania and Schr\"oter, Isabel Carolin and MacDorman, Karl F. and Teufel, Martin and B\"auerle, Alexander, 2024
- Fundamentals for a Child-Oriented Approach to Data Processing (opens in a new tab) · Data Protection Commission, 2021
- Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (opens in a new tab) · Federal Bureau of Investigation, 2024
- Scammers use AI to enhance their family emergency schemes (opens in a new tab) · Puig, Alvaro, 2023
- Launch of Hotline to Report Intimate Image Abuse (IIA) to Garda\'i (6/9/21) (opens in a new tab) · An Garda S\'ioch\'ana, 2021
- Fraud (opens in a new tab) · An Garda S\'ioch\'ana, 2025
- Online Child Exploitation (opens in a new tab) · An Garda S\'ioch\'ana, 2025
- Age requirements on Google Accounts (opens in a new tab) · Google, 2026
- Guide your child's Gemini Apps experience (opens in a new tab) · Google, 2026
- LCQ9: Combating frauds involving deepfake (opens in a new tab) · Government of the Hong Kong Special Administrative Region, 2024
- FAQs (opens in a new tab) · Irish Internet Hotline, 2026
- Intimate image abuse (IIA) (opens in a new tab) · Irish Internet Hotline, 2026
- Report Illegal Content Online (opens in a new tab) · Irish Internet Hotline, 2026
- What to Report Online (opens in a new tab) · Irish Internet Hotline, 2026
- Age appropriate design: a code of practice for online services (opens in a new tab) · Information Commissioner's Office, 2026
- Harassment, Harmful Communications and Related Offences Act 2020 (Act No. 32 of 2020) (opens in a new tab) · Houses of the Oireachtas, 2020
- Navigating the internet safely campaign (opens in a new tab) · Internet Watch Foundation, 2026
- Report online child sexual abuse imagery or 'child pornography' (opens in a new tab) · Internet Watch Foundation, 2026
- New online reporting service for victims of intimate image abuse (opens in a new tab) · Jacques, Paul, 2026
- Warning: Humans cannot reliably detect speech deepfakes (opens in a new tab) · Mai, Kimberly T. and Bray, Sergi and Davies, Toby and Griffin, Lewis D., 2023
- Be Media Smart: Tips (opens in a new tab) · Media Literacy Ireland, 2026
- How to spot a scam email, text message or call (opens in a new tab) · National Cyber Security Centre (UK), 2021
- ChatGPT and large language models: what's the risk? (opens in a new tab) · Chismon, Dave and Paul J, 2023
- Thinking about the security of AI systems (opens in a new tab) · Martin R, 2023
- The near-term impact of AI on the cyber threat (opens in a new tab) · National Cyber Security Centre (UK), 2024
- Thinking carefully before adopting agentic AI (opens in a new tab) · Martin R and Kate S, 2026
- How to keep your online accounts secure: Advice from the National Cyber Security Centre (opens in a new tab) · National Cyber Security Centre (Ireland), 2025
- March 20 ChatGPT outage: Here's what happened (opens in a new tab) · OpenAI, 2023
- Europe Terms of Use (opens in a new tab) · OpenAI, 2026
- Sharing conversations and scheduled tasks in ChatGPT (opens in a new tab) · OpenAI, 2026
- Banking Fraud (opens in a new tab) · Police Scotland, 2026
- Deepfakes (opens in a new tab) · Police Scotland, 2026
- Warning against fraud after victim loses \pounds250,000 in Ards and North Down to online criminals (opens in a new tab) · Police Service of Northern Ireland, 2026
- How to get in touch with the Revenge Porn Helpline (opens in a new tab) · Revenge Porn Helpline, 2026
- Advice for under 18s who have had private sexual content shared online (opens in a new tab) · Revenge Porn Helpline, 2026
- OpenAI Is Pulling Shared ChatGPT Chats From Google Search (opens in a new tab) · Southern, Matt G., 2025
- Reporting fraud (opens in a new tab) · UK Government, 2026