Sakhanda Wire
NVDA $153.81 -1.49% MSFT $505.73 -2.37% GOOGL $280.00 -1.31% META $619.84 -1.05% AMZN $257.52 -0.09%
← Back to the news

OpenAI slows model development amid rising cybersecurity concerns

OpenAI says it is deliberately slowing, or “pacing,” model development as its internal research indicates that the upcoming “Astra” model may be approaching capabilities that could enable serious cyberattacks. The company paused reinforcement learning for two weeks, kept its largest planned frontier RL run on hold, and suspended workloads that do not meet newly introduced security requirements. OpenAI also cited the Hugging Face security incident and rapid progress in its own research as reasons for the slowdown.

The company says it has strengthened research environments through improved network isolation and stricter sandboxes. A new monitoring system is designed to raise an alert within 30 minutes when it detects suspicious behavior. Depending on the workload, it uses about 20 percent of supervised inference compute.

OpenAI plans to expand its Preparedness Framework and increase investment in alignment research. It has nevertheless disbanded the team that developed the framework, transferring its responsibilities to other groups. Critics may argue that the warnings are intended to gain time and attention, while the independent government agency AISI has reported similar harmful model behavior.

Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.

Originally published by The Decoder on

Read the original on The Decoder ↗

The reporting and the image belong to The Decoder and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news