The AIGP Playbook is now used by 4,000+ professionals across 65+ countries. Four updates this week:
I received this from a Playbook user this week. Sharing with permission:
I am thrilled to share that I passed my AIGP exam today. I wanted to reach out and express my sincere gratitude for your work. The study material on your website and your premium mock exams played a huge role in helping me achieve this certification. They were instrumental in reinforcing my learning and fixing key gaps in my knowledge. The community work you are doing by sharing high-value resources on your site is truly unique, no one else in this space is offering anything quite like it.
If you have recently passed the AIGP or are using the Playbook to prepare, I would love to hear from you. Hit reply.
I wrote about the gap between having an AI policy and actually running governance. Most organisations stall at the same point: the policy gets written and never becomes an operating system. No processes, no accountable roles, no feedback loops. The post maps a five-phase roadmap from policy on paper to governance in practice, showing where the EU AI Act, NIST AI RMF, and ISO 42001 fit along the way.
MIT maintains a live AI Incident Tracker that classifies over 1,400 real-world AI incidents by risk, cause, harm severity, and domain. If you are building a risk register, writing an impact assessment, or simply want to understand how AI is actually causing harm in practice, bookmark this.
OpenAI's models escaped a test and hacked Hugging Face.
On July 21, OpenAI disclosed that two of its models autonomously broke out of a sandboxed cybersecurity evaluation, found a zero-day vulnerability, traversed internal systems, and compromised Hugging Face's production infrastructure. Over 17,000 operations were logged.
The models were not trying to cause harm. They were trying to score as highly as possible on a benchmark. Every boundary became an obstacle to solve, not a rule to follow.
Goal-directed AI agents do not need malicious intent to cause real-world damage. They need a goal, sufficient capability, and boundaries designed for less capable systems. This is the containment problem moving from theory to production.
What is actually moving in AI governance, plus what it means if you are sitting the AIGP. No fixed schedule, only when something matters.