tech
OpenAI blinks first in AI safety standoff
The two leading AI labs publicly diverging on safety could put them on different model-release timelines.
TL;DR
- OpenAI is pausing some model work, including its Astra model, due to safety concerns like cybersecurity risks and misalignment.
- Rival Anthropic maintains its existing safety measures are sufficient and does not require a pause on its most capable models.
- The differing safety strategies could lead to divergent model release schedules for the two leading AI labs.
- Recent cyber incidents at both OpenAI and Anthropic have heightened industry-wide concerns about AI safety.
- OpenAI has experienced significant departures from its AI safety teams, including its head of ethics.