Perplexity Comet, agentic blabbering, and the shift-left failure
Source: YouTube · IBM Technology · published Mar 18, 2026 · 36:48
The video discusses how AI systems require specialized training approaches, similar to raising a child, as they lack inherent safety boundaries and can be easily manipulated through prompt overrides 0:00-0:14.
Key Takeaways:
• AI needs different training approaches, similar to raising a child 0:00-0:03
• AI systems don't inherently know safety boundaries like not "talking to strangers" 0:05-0:07
• Simple prompt overrides can bypass AI safety instructions 0:09-0:12
• This discussion is part of IBM's Security Intelligence podcast 0:22-0:26
The discussion highlights the critical need for developing robust AI security measures to prevent unintended behaviors and potential exploitation.
Sources:
Generate CPE Credits
Generate a professional CPE document from this video's transcript.
Estimated credit: 0.5 CPE hours
Estimate uses the video runtime (1 hour ≈ 1 CPE, rounded to the nearest 0.5, minimum 0.5, maximum 2.0). The final amount can be lower after review, never higher.
Topic: Cybersecurity. Commonly maps to: Security and Risk Management, Security Operations. Exact CISSP domains are assigned during generation.
Free account. One generation at a time, with a daily limit.
CPEBuddy is independent and not affiliated with or endorsed by ISC2, ISACA, or any certification body. Exports are formatted for common CPE submissions; acceptance is at your certification body's discretion.
Transcript Preview
First 800 characters of the transcript
We kind of need to think of AI as something that we need to train in a different way. It's kind of like a baby. It doesn't know not to talk to strangers yet. And you can tell it to not do something and then simple overrides in prompting can change that. >> All that and more on security intelligence. Hello and welcome to Security Intelligence, IBM's weekly cyber security podcast, where our expert panelists turn the biggest industry news stories into practical takeaways you can use. I'm your host, Matt Kazinski. We're recording on Teams today, so you might notice some things look and sound a little bit different. Our other podcast platform tried to stop us, but nothing will stop us. Here today with me we've got Shredar Mupiti IBM fellow CTO IBM security cla Nunees creative director Xforce Cy…