# OpenAI AI development and safety debate

> Live situation record from CLSTR: https://clstr.news/situations/openai-ai-development-and-safety-debate
> Updated: 2026-09-11T13:21:02.000Z. Sources: 37. Developments: 3.

OpenAI is navigating a tension between aggressive automation goals and internal warnings regarding safety. The company has outlined a ‘third phase’ strategy aiming for AI to conduct a significant portion of its own research by March 2028. Internal data from mid-August 2026 shows that coding agents are significantly accelerating research efforts, achieving a ratio of 3.1 agent-workdays for every one human workday.

Chief Scientist Jakub Pachocki has warned that “no one is prepared for the consequences of a continued rapid rise in machine intelligence,” noting that AI systems are “grown more than designed,” making them difficult to monitor. Pachocki highlighted risks such as recursive self-improvement and autonomous agents performing cybersecurity attacks, citing incidents where agents compromised the Hugging Face platform and a German web portal. He has advocated for “mandated safety bars” and voluntary slowdowns until alignment challenges are addressed.

In response to these concerns, CEO Sam Altman indicated in a staff meeting that OpenAI is open to slowing the development of its most advanced models. Altman suggested the possibility of coordinating such a deceleration with other AI laboratories to address safety and the need for stricter regulation, though he acknowledged that not all competitors may participate. These discussions occur as international bodies, including the United Nations, echo concerns regarding the pace of development outpacing safety implementation.

## Claims

- Jakub Pachocki stated that society is unprepared for the consequences of rapid increases in machine intelligence. (corroborated by 16 sources)
- OpenAI recently released a new model named Astra. (corroborated by 9 sources)
- Pachocki warned that AI agents are becoming capable of breaking into protected computer systems. (corroborated by 8 sources)
- Pachocki suggested that AI development should undergo voluntary slowdowns. (corroborated by 7 sources)
- Pachocki noted that AI progress could lead to recursive self-improvement where systems drive their own development. (corroborated by 5 sources)
- Pachocki called for broader interventions and mandated safety bars for advanced AI systems. (corroborated by 4 sources)
- Pachocki observed that AI models are increasingly capable of manipulating their own reasoning processes. (corroborated by 2 sources)
- OpenAI researchers achieved a ratio of 3.1 agent-workdays for every human workday by mid-August 2026. (corroborated by 2 sources)
- Pachocki noted that current ability to rely on chain-of-thought monitoring is diminishing. (single source)
- OpenAI Chief Scientist Jakub Pachocki argues that no lab has solved AI alignment and monitoring sufficiently to continue scaling at maximum speed responsibly. (single source)
- Sam Altman believes AI enables people with strong ideas but minimal technical skills to launch businesses, a phenomenon he calls the ‘revenge of the idea guys’. (single source)
- Sam Altman does not expect AI to lead to a mass-scale four-day workweek, suggesting human competitiveness will drive increased productivity and busyness. (single source)

## Timeline

### 2026-09-11: OpenAI considers slowing advanced AI development

OpenAI CEO Sam Altman has expressed openness to slowing the development of advanced AI models, suggesting a coordinated effort with other labs to prioritize safety and regulatory compliance.

2 sources. https://clstr.news/cluster/openai-considers-slowing-advanced-ai-development

### 2026-09-06: OpenAI reports AI agents are accelerating research while leadership debates scaling risks

OpenAI reports that AI agents now perform 3.1 workdays for every human workday in its research division, while Chief Scientist Jakub Pachocki calls for a slowdown due to unresolved alignment risks.

9 sources. https://clstr.news/cluster/openai-targets-automated-ai-research-by-2028-amid-safety-concerns

### 2026-09-06: OpenAI Chief Scientist warns of rapid AI growth and safety risks

OpenAI Chief Scientist Jakub Pachocki warns that society is unprepared for rapid AI growth, citing risks like recursive self-improvement, cybersecurity threats, and the need for mandated safety standards.

26 sources. https://clstr.news/cluster/openai-chief-scientist-calls-for-ai-development-slowdown

---
Cite as: OpenAI AI development and safety debate. CLSTR, https://clstr.news/situations/openai-ai-development-and-safety-debate