Note: my apologies for the absence in two weeks. I caught some incidents after returning to Vietnam from ICLR, 2025 Singapore. This week, I will try to upload more posts to help you get updated.
Highlights
Gemini SafeguardsAnthropic: ASL-3 deployed.
Gemini: Adaptive attack defense.Thales: Quantum threat rising.NIST: AI data security guide.
Dive Deep
1. Gemini Security Safeguard Google DeepMind
Automated red-teaming increased protection.
Model hardening boosts resilience.
Adaptive attacks tested defenses.
AI-specific security spending up.
Defense-in-depth approach adopted.
Headsup: Considered automated red team to boost security.
2. Claude 4 Safety Level 3 Anthropic
Activated AI Safety Level -3.
CBPR: Chemical, Biological, Radiological and Nuclear.
Deployed Claude Opus 4.
Not yet passed the ASL-3, but need ASL-4.
Continued improvement with more knowledge.
Headsup: New measures and the rational behind.
3. Thales: AI and Quantum Threats Thales
56% cite enterprise breach.
63% fear encryption compromise.
58% worry about “decrypt later”.
60% prototyping PQC solutions.
Challenge in innovation vs security.
Headsup: Assess quantum readiness; prioritize PQC prototyping now.
4. Breach: 184M Logins and Passwords ZDNET
184M credentials exposed.
Data unencrypted, publicly accessible.
Infostealer malware suspected.
Facebook, Google, Microsoft impacted.
Metadata poses counterintelligence risk.
Headsup: Immediately enforce password resets & MFA.
5. NIST: AI Data Security NIST
Best practices for securing data.
Focused on 6 stages of data life cycle.
Risks: supply chain, poisoned, data drift.
Strategic overview, not technical documents.
Joint statements of top agencies.
Headsup: checkout our latest review for more.
Notice: Y2 GenAI Safety and Security is on GumRoad and Amazon with paperback.
Share this post