> Markdown version of [/videos/1316-fighting-fraud-with-an-ai-grandma-ben-hopkins-and-morten-legarth-from-faith-vccp?t=2039](https://www.wearedevelopers.com/videos/1316-fighting-fraud-with-an-ai-grandma-ben-hopkins-and-morten-legarth-from-faith-vccp?t=2039). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Fighting Fraud with an AI Grandma - Ben Hopkins and Morten Legarth from faith @ VCCP

Ben Hopkins and Morten Legarth built an AI grandmother to trap scammers. Her elderly persona cleverly masks LLM latency. Discover how this safely harvests actionable threat intelligence.

- **Speakers:** Ben Hopkins, Morten Legarth
- **Event:** Coffee With Developers
- **Published:** March 28, 2025
- **Duration:** 38:09
- **URL:** https://www.wearedevelopers.com/videos/1316-fighting-fraud-with-an-ai-grandma-ben-hopkins-and-morten-legarth-from-faith-vccp

## Summary

In an effort to combat the growing epidemic of phone fraud, Ben Hopkins and Morten Legarth from VCCP's AI division, "faith," helped develop an innovative conversational AI known as "Daisy." Created for the telecommunications brand O2, Daisy is an "AI Grandma" designed to answer calls from scammers and keep them engaged in meandering, realistic conversations. By moving away from traditional, passive educational awareness campaigns toward proactive "story-doing," this initiative ties up malicious actors' time, preventing them from defrauding real victims while organically boosting public awareness of fraud reporting tools.

The project highlights the critical intersection of technical engineering and creative world-building. To make the AI believable, the team had to look beyond standard prompt engineering and narrative voice libraries, such as those initially found on platforms like 11Labs. By mapping the persona of an elderly woman, the team successfully masked inevitable LLM latency; the natural pauses of an older character reasoning or mishearing seamlessly disguised algorithmic processing delays. Furthermore, training the voice model on hours of genuine, unstructured conversations with a colleague's grandmother provided the conversational cadence necessary to fool highly motivated threat actors during live calls.

Beyond its immediate success in social engineering defense, Daisy offers profound insights into modern AI application development. The creators emphasize that approaching generative AI with an optimistic, responsible mindset can accelerate human creativity rather than stifle it. Additionally, running fine-tuned models on consumer-grade local GPUs mitigated the exorbitant energy consumption typically associated with continuous cloud-based AI generation. Looking forward, deploying autonomous agents like Daisy opens new pathways in cybersecurity; because AI bots do not possess personal data rights, organizations can safely record and analyze scammer interactions to harvest threat intelligence without violating GDPR or compromising real customer privacy.

**Keywords:** conversational AI development, generative AI character design, voice cloning technology, LLM latency optimization, phone scam prevention, faith at VCCP, AI grandmother daisy, cybersecurity threat intelligence, story-doing marketing, natural language processing, consumer-grade GPU AI deployment, scam baiting automation, LLM fine-tuning, GDPR compliant data capture, AI energy consumption

## Chapters

1. **Establishing an optimistic AI viewpoint for software agencies** (00:01) — Embracing generative tools responsibly acts as a creative accelerator instead of a threat to established teams.
1. **Building an AI persona to combat telecom scammers** (03:44) — Deploying an automated conversational agent creates a disruptive mechanism that wastes bad actors' time while driving brand awareness.
1. **Prototyping conversational models with small technical teams** (06:56) — Creating the prototype application required blending computer science workflows directly with narrative character profile building.
1. **Masking system latency through strategic character design** (09:15) — An elderly persona naturally turns slow programmatic response times and simulated communication disruptions into believable interaction traits.
1. **Addressing privacy considerations when targeting digital cybercriminals** (13:26) — Legal evaluations determined that preventing active criminal behavior justified training conversational models on unfiltered scamming data.
1. **Keeping language model prompt design lean for speed** (15:22) — Streamlining persona prompts prevents overcomplicated context parsing from introducing unwanted conversational delays at runtime.
1. **Creating believable synthetic voices for conversational AI** (17:34) — Cloning authentic dialogue from untrained voice talent overcomes the unnatural theatrical performance style typical of commercial audio platforms.
1. **Testing the conversational agent against live telecommunication exploiters** (22:34) — Running the initial software successfully secured a bad actor in a smooth continuous interaction loop for fifteen minutes.
1. **Evaluating the hardware footprint and energy costs of audio** (26:59) — Deploying fine-tuned voice models locally on consumer hardware sidesteps the prohibitive power demands of large multimodal image generation systems.
1. **Measuring campaign effectiveness and telecommunication incident reporting** (30:52) — Beyond exhausting hacking resources proactively, releasing the conversational tool drove substantial increases in protective telemetry reporting.
1. **Demonstrating automated defense strategies at prominent cybersecurity conferences** (33:59) — Exposing the automated system to security red teams demonstrates robust methodologies for assessing infrastructure threats without jeopardizing real consumers.

## Related Moments

- [Real-world voice cloning and targeted fraud cases](https://www.wearedevelopers.com/videos/770-skynet-wants-your-passwords-the-role-of-ai-in-automating-social-engineering) (from "Skynet wants your Passwords! The Role of AI in Automating Social Engineering")
- [Enhancing social engineering and phishing with generative AI](https://www.wearedevelopers.com/videos/926-wwc24-chris-wysopal-helmut-reisinger-and-johannes-steger-fighting-digital-threats-in-the-age-of-ai) (from "WWC24 - Chris Wysopal, Helmut Reisinger and Johannes Steger - Fighting Digital Threats in the Age of AI")
- [Addressing the authenticity of AI translated video content](https://www.wearedevelopers.com/videos/1819-wearedevelopers-live-yes-css-can-do-that) (from "WeAreDevelopers LIVE - Yes, CSS Can Do That!")
- [Demonstrating a vulnerable AI sales agent application](https://www.wearedevelopers.com/videos/1563-prompt-injection-poisoning-more-the-dark-side-of-llms) (from "Prompt Injection, Poisoning & More: The Dark Side of LLMs")
- [Automating compromised roles through AI agents and backup scammers](https://www.wearedevelopers.com/videos/100057-synthetic-insiders-the-new-ai-risk-to-your-org) (from "Synthetic Insiders: The New AI Risk to Your Org")
- [Scaling malicious tasks with autonomous AI agents](https://www.wearedevelopers.com/videos/770-skynet-wants-your-passwords-the-role-of-ai-in-automating-social-engineering) (from "Skynet wants your Passwords! The Role of AI in Automating Social Engineering")

## Related Articles

- [WWC24 Talk - Scott Hanselman - AI: Superhero or Supervillain?](https://www.wearedevelopers.com/magazine/469-wwc24-talk-scott-hanselman-ai-superhero-or-supervillain)
- [Dev Digest 135 - AI is Doomed](https://www.wearedevelopers.com/magazine/480-dev-digest-135-ai-is-doomed)
- [Exploring AI: Opportunities and Risks for Developers](https://www.wearedevelopers.com/magazine/522-exploring-ai-opportunities-and-risks-for-developers)
- [Coffee with Developers - Maria Apazoglou - Making AI understandable for all in production](https://www.wearedevelopers.com/magazine/475-coffee-with-developers-maria-apazoglou-making-ai-understandable-for-all-in-production)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Principal Field Architect - AI Agents](https://www.wearedevelopers.com/jobs/ext/1442858-principal-field-architect-ai-agents) at **Twilio**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**
- [Twilio's next Senior Principal Field Architect - AI Agents](https://www.wearedevelopers.com/jobs/ext/1487390-twilio-s-next-senior-principal-field-architect-ai-agents) at **Twilio**
- [AI Operations Manager (all genders)](https://www.wearedevelopers.com/jobs/48263-ai-operations-manager-all-genders) at **envelio**
- [Senior Software Engineer, Fraud](https://www.wearedevelopers.com/jobs/ext/1280398-senior-software-engineer-fraud) at **Twilio**