AI models are powerful, but they don’t forget. And that's a problem.

They hallucinate. They inherit bias. They absorb sensitive data. And once they’re trained, fixing those issues is painfully expensive. Retraining takes weeks and maybe tens of millions of dollars. And any guardrails the AI company puts up are brittle.

What if you could perform surgery on the model itself?

In this episode of TechFirst, John Koetsier sits down with Ben Luria, co-founder of Hirundo, to explore machine unlearning, a new approach that selectively removes unwanted data, behaviors, and vulnerabilities from trained AI systems.

Hirundo claims it can:

• Cut hallucinations in half

• Massively reduce bias

• Reduce successful prompt injection attacks by over 90%

• Do it in under an hour on a single GPU

• Preserve benchmark performance

Instead of adding more guardrails, machine unlearning works inside the model, identifying problematic weights, isolating behavioral vectors, and surgically removing risks without degrading quality.

If AI is going mainstream in enterprises, it needs a remediation layer. Is machine unlearning the missing piece?

⸻

Guest

Ben Luria

Co-Founder, HirundoNhir

https://www.hirundo.io

⸻

Topics Covered

• Why AI models “can’t forget”

• The difference between hallucinations and inaccuracies

• Why guardrails aren’t enough

• How prompt injection works — and how to reduce it

• Removing PII and noncompliant training data

• AI security at the model level

• Why machine unlearning could become standard by 2030

⸻

If you’re building, deploying, or investing in AI, this is a conversation you can’t miss.

👉 Subscribe for more deep dives into AI, innovation, and the future of tech:

https://techfirst.substack.com

⸻

⏱ Chapters

00:00 – Why We Need Machine Unlearning

01:12 – What Is Machine Unlearning?

03:40 – Why AI Can’t “Forget” (The Pink Elephant Problem)

06:15 – Guardrails vs True Model Remediation

09:05 – The Wild West of AI Data & Legal Risk

11:20 – How Machine Unlearning Works (Detection, Isolation, Remediation)

16:10 – Performing “Neurosurgery” on LLMs

19:30 – Hallucinations vs Inaccuracies Explained

23:45 – Reducing Prompt Injection by 90%

28:30 – Working with AI Labs & Enterprises

32:00 – Will Unlearning Become Standard by 2030?

34:15 – Final Thoughts