Daily · Wednesday, August 19, 2026
OpenAI Halts Astra for Safety; Block Open-Sources Berd Workspace
- Open AI roles
- 53 on the weekup
- Created month to date
- 117,917 Modelled
- Employers tracked
- 166
Safety is slowing one headline project even as agent tooling speeds up. OpenAI paused training for its Astra model to install new cybersecurity-focused protocols. Meanwhile Block open-sourced Berd, a cross-model AI agent workspace on GitHub under Apache 2.0, potentially broadening access to agent workflows. Across the 166 employers we track, 3,624 AI roles were open as of 2026-08-19, up 53 on the week. Our model puts new AI jobs created month-to-date at roughly 117,917. Yet only 39 of those are safety roles, and hiring remains concentrated in engineering and data. Practitioners are touting capability leaps even as they talk about spending ceilings and a trust crisis; vendor hiring looks steady, but evaluation and security work is thin. The result is a widening gap between caution at the model frontier and where headcount is going. For people doing the work, this points to two paths: more demand to build, integrate and operate agents across stacks, and rising expectations that engineers shoulder security, testing and monitoring without dedicated safety teams.
The stories
- TOOL_LAUNCH
OpenAI Halts Astra Model Training to Enhance Safety Protocols
OpenAI has paused training for its upcoming Astra model to implement new safety protocols addressing cybersecurity risks.
Why it mattersA training pause at a major lab shifts attention to cybersecurity and evaluation expertise inside model development teams.
- TOOL_LAUNCH
Block Open-Sources Berd, a Cross-Model AI Agent Workspace
Block has open-sourced Berd, a cross-model AI agent workspace, available on GitHub under the Apache 2.0 license.
Why it mattersBy open-sourcing a cross-model agent workspace, Block could expand demand for integration, orchestration, and automation skills.
- ANALYSIS
Safety alarms grow; only 39 safety roles in 3,624 AI openings
This week’s posts flag rising alignment risk and rapid adoption, while our tracker shows hiring remains concentrated in engineering and data, with few evaluation and safety roles.
Why it mattersDespite rising risk talk, our tracker shows just 39 safety roles, revealing a stark hiring imbalance.
- ANALYSIS
Engineering hires dwarf safety as hype swirls around AI
Practitioners tout capability leaps, spending ceilings and trust crises; our 3,660-role tracker shows engineering demand surging, safety thin, and vendor hiring steady.
Why it mattersThe analysis highlights engineering demand outpacing safety work while vendor hiring remains steady amid capability hype and trust concerns.