From the source

Safety and alignment in an era of long-horizon models

Safety or policy

OpenAI published a post sharing lessons from deploying long-running AI models, focusing on safety risks, observed failures, and improved safeguards.

From the source
OpenAI shares lessons from deploying long-running AI models, highlighting new safety risks, observed failures, and improved safeguards through iterative deployment.
openai.com

Other stories

SentinelOne cuts malware analysis from weeks to hours
SentinelOne uses OpenAI Daybreak to speed reverse engineering from weeks to hours.
OpenAI· 14 Oct
Amp puts AI agents to work beyond code with OpenAI models
OpenAI· 14 Oct
How Block is building AI-powered security with OpenAI technology
Block uses OpenAI Daybreak to speed security operations
OpenAI· 14 Oct
How Third Moment Research runs 10,000 agents for 36+ hours

Announced: 20 Jul 2026 10:00 UTC
Detected: 03 Aug 2026 21:11 UTC (+14d after)

Visit source ↗openai.com

OpenAI

· 14 Oct

The EU’s next chapter in AI is agentic

OpenAI reports agentic AI use now accounts for over half of EU enterprise output tokens.

OpenAI· 14 OctResearch paper

Sophos cuts threat investigation time by 96% with OpenAI Daybreak

Sophos uses OpenAI Daybreak to automate 52% of MDR cases, cutting response time to 89 seconds.

OpenAI· 9 Oct

Safety and alignment in an era of long-horizon models

From the source

SentinelOne cuts malware analysis from weeks to hours

Amp puts AI agents to work beyond code with OpenAI models

How Block is building AI-powered security with OpenAI technology

How Third Moment Research runs 10,000 agents for 36+ hours

The EU’s next chapter in AI is agentic

Sophos cuts threat investigation time by 96% with OpenAI Daybreak