Software engineering makes up ~50% of agentic tool calls on our API, but we see emerging use in other industries. As the frontier of risk and autonomy expands, post-deployment monitoring becomes essential. We encourage other model developers to extend this research.
SAFETY
-
Claude Code Enhances AI Safety Through Uncertainty Recognition
By
–
Claude Code also encourages oversight by stopping to ask questions. On complex tasks, Claude Code pauses for clarification more than twice as often as humans interrupt it. Training models to recognize uncertainty is an important, under-appreciated safety property.
-
Agent API Risk: Security, Finance, and Production Deployment Frontiers
By
–
Most agent actions on our API are low risk. 73% of tool calls appear to have a human in the loop, and only 0.8% are irreversible. But at the frontier, we see agents acting on security systems, financial transactions, and production deployments (though some may be evals).
-
Real-world agent deployment: safety implications across contexts
By
–
Agents are already being deployed across contexts that range from e-mail triage to cybersecurity research. Understanding this spectrum is critical for safe deployment, yet we know surprisingly little about how people actually use agents in the real world.
-
Measuring Agentic Security Capabilities with Smart Contracts
By
–
measuring agentic security capabilities with smart contracts:
-
AI Agents Security: Why Undo Functionality Matters for Organizations
By
–
On the AI Security Podcast, @RubrikInc GM of AI Dev Rishi covers the potential damage that can be caused by AI agents and why the “undo” important will be critical for organizations going through an #AI transformation 👉 go.rbrk.co/xobwcp
→ View original post on X — @predibase, 2026-02-18 18:15 UTC
-
PolyAI voice realism could make real people sound like AI
By
–
Great demo video
— AI Breakfast (@AiBreakfast) 18 février 2026
Millions of people have interacted with a PolyAI voice and didn't even realize it wasn't a person
This tech will get so good that the script will flip, and you'll start to think real people on the phone are AI, because it'll be a safer assumption at that point https://t.co/fYtFaSAuGcGreat demo video Millions of people have interacted with a PolyAI voice and didn't even realize it wasn't a person This tech will get so good that the script will flip, and you'll start to think real people on the phone are AI, because it'll be a safer assumption at that point
-
Pennsylvania releases 389-page AI adoption recommendations
By
–
Commonwealth of Pennsylvania releases comprehensive recommendations on AI adoption and use: https://
linkedin.com/posts/ctkraft_
pa-ai-recommendations-ugcPost-7428286689254563840-kZBf
… (Download 389-page PDF report ) -
Smart Metrics and Governance: Curbing Metric Gaming in AI
By
–
Smart metrics don't just measure performance, they shape behavior. If we're deploying #AI across the enterprise, we need guardrails against metric gaming from day one. Governance isn't a brake on innovation, it's what makes transformation durable. Especially as agentic systems
-
AI Agents Identity Problem Not Security Problem
By
–
AI agents don't have a security problem.
— Akshay 🚀 (@akshay_pachaar) 18 février 2026
They have an identity problem.
Here's what I mean:
You've deployed 47 agent workflows across your infrastructure.
At 3 am, an agent modifies a production database. You check your logs. 47 agents had access. They all used the same… pic.twitter.com/TOsOEE3PbtAI agents don't have a security problem. They have an identity problem. Here's what I mean: You've deployed 47 agent workflows across your infrastructure. At 3 am, an agent modifies a production database. You check your logs. 47 agents had access. They all used the same