
🎙️ EP 277: Bypassing AI Guardrails in Minutes & The MAI-Image-2.5 Power Drop
Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode!
The core safety guardrails of Meta’s Llama 3.3 and Google’s Gemma models were stripped away in under ten minutes using a standard laptop and a free GitHub tool called "Heretic." We're parsing the explosive Financial Times investigation on "abliteration" and what this means for the open-source vs. closed-source AI war. We also look at the newly released MAI-Image-2.5 from Microsoft's MAI team, which just stormed the global Arena leaderboard at No. 3.
In this episode, we cover:
Inside the…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.