GPT 5.5: The System Card

GPT 5.5: The System Card

From Don't Worry About the Vase Podcast by Podcast for Zvi's blog, Don't Worry About the Vase Podcast

April 27, 2026 · 27 min

About this episode

This episode discusses the system card for GPT 5.5, covering various aspects of AI safety and evaluation.

The Don’t Worry About the Vase Podcast is a listener-supported podcast. To receive new posts and support the cost of creation, consider becoming a free or paid subscriber. * 00:00 - Introduction * 02:48 - Pro Versus Proxy * 03:10 - Disallowed Content (3.1) * 05:29 - Don’t Delete Data (3.3) * 06:30 - Confirmation Confirmation (3.4) * 06:56 - Jailbreaks (4.1) * 07:07 - Prompt Injections (4.2) * 08:38 - Health (5) * 09:01 - Hallucinations (6) * 10:16 - Alignment (7) * 14:48 - Bias Evaluation (8) * 15:16 - Preparedness (9) * 16:17 - Bio (9.1.1) * 18:30 - Cybersecurity (9.1.2) * 20:53 - Self-Improvement (9.1.3) * 21:50 - Sandbagging (9.2) * 22:53 - Safeguards (9.3) * 24:52 - What About Model Welfare? * 25:42 - Would This Have Identified A Problem? https://open.substack.com/pub/thezvi/p/gpt-55-the-system-card?utm_campaign=post-expanded-share&utm_medium=web Get full access to DWAtV Podcast at dwatvpodcast.substack.com/subscribe

Topics covered

  • AI
  • technology
  • data privacy
  • cybersecurity
  • model evaluation
  • self-improvement

Keywords

  • GPT 5.5
  • system card
  • data
  • jailbreaks
  • prompt injections
  • alignment
  • bias evaluation
  • cybersecurity

Mentioned in this episode

Organizations: Don't Worry About the Vase Podcast

Products: GPT 5.5

More episodes of Don't Worry About the Vase Podcast

Explore listener stats, chart rankings, contacts and more on the Don't Worry About the Vase Podcast podcast page.