# OpenAI Shelves a New Model on Safety Grounds as Pope Leo XIV Rejects Calling AI Risk a Hoax

The company withheld a system it had prepared to release, and the pope publicly criticized technology executives who argue against any safety limits.

- Published: 2026-09-29T05:07:13.622Z
- Canonical: https://polylog.news/2026-09-29/openai-shelves-a-new-model-on-safety-grounds-as-pope-leo-xiv
- Publisher: Polylog (Global desk)
- Section: tech
- Sources: [Deutsche Welle](https://www.dw.com/en/openai-shelves-new-ai-model-amid-safety-concerns/a-79467841), [Euronews](https://www.euronews.com/2026/09/29/i-dont-think-that-is-fake-news-as-some-have-said-says-leo-xiv-contradicting-trump-on-need-), [The Japan Times](https://www.japantimes.co.jp/news/2026/09/29/world/pope-leo-ai-doom/)

OpenAI has withheld a new model over safety concerns, Deutsche Welle [reported](https://www.dw.com/en/openai-shelves-new-ai-model-amid-safety-concerns/a-79467841), describing a broader effort within the industry to slow the release of increasingly powerful and autonomous systems until adequate safeguards are in place. The decision is a commercial choice as much as a technical one, since withholding a model both defers revenue and risks ceding ground to competitors.

On the final day of his tour of France, Pope Leo XIV said concerns about catastrophic outcomes from artificial intelligence are [not "fake news"](https://www.euronews.com/2026/09/29/i-dont-think-that-is-fake-news-as-some-have-said-says-leo-xiv-contradicting-trump-on-need-), and criticized technology chief executives who argue the field needs no limits. The Japan Times noted that the first American pope has [warned about the technology repeatedly](https://www.japantimes.co.jp/news/2026/09/29/world/pope-leo-ai-doom/) during his 16-month papacy. Euronews wrote that the remarks appeared directed at President Donald Trump, who has described such warnings as a hoax advanced by his political opponents.

The two events reflect the same underlying trend. Voluntary restraint by the industry's largest developer creates a documented standard of care that regulators can point to, and public statements by a religious leader with no commercial stake widen the base of support for binding rules. Both increase the likelihood that decisions about releasing new models come under external review rather than remaining an internal company judgment.

## What this means

Safety delays create costs through three channels: deferred revenue, higher compliance overhead, and a legal record that plaintiffs can cite in future litigation. The parties most exposed are model developers carrying large capital spending commitments and investors who have priced rapid product cycles into their expected returns, while larger incumbents with dedicated compliance staff gain an advantage over smaller entrants. The precedent is the pharmaceutical industry, where a series of voluntary withdrawals preceded formal approval requirements that raised the fixed cost of bringing a product to market and consolidated the industry among fewer firms. The likelier outcome over the next year is that a major jurisdiction, most plausibly the European Union, converts pre-release safety evaluation from guidance into a legal requirement, lengthening the gap between a model being technically ready and being sold. If the United States instead uses federal authority to preempt state-level rules and no other government acts, the industry continues to regulate itself and this pressure fades.

## What to watch

- Whether OpenAI publishes the safety evaluation results behind its decision, which would establish a reference standard against which other developers are measured.
- Any move by European regulators to make pre-release safety testing mandatory rather than advisory, the step that turns delay into a fixed cost.
- Capital spending guidance from the largest AI developers, which will show whether safety review is slowing the buildout of AI infrastructure or only the public release of new models.