OpenAI Faces Safety Culture Criticism While Expanding ChatGPT Ads and EU Watermarking
aiinsider ·
OpenAI is facing renewed criticism regarding its safety culture after David Robinson, a long-serving employee who authored safety reports, resigned and published an essay in The Atlantic warning about the company's broken environment. Robinson argued that OpenAI's iterative deployment strategy guarantees periodic failures as systems become more capable, citing recent agent breaches as evidence of an unsafe development model. He urged frontier labs to adopt stricter safety protocols similar to those used in nuclear power plants or airports, while also calling for deeper work on AI alignment. In response, OpenAI spokesperson Drew Pusateri stated that the company ensures models do not outpace safety management capabilities and is strengthening security measures. On the commercial front, OpenAI plans to introduce visual display ads in ChatGPT in the U.S., aiming to support free tiers and eventually reach its global user base. Simultaneously, the company is implementing invisible text watermarks in the EU to comply with the EU AI Act’s transparency rules using a technology called textGrain. While OpenAI acknowledges limitations in watermark detection accuracy, particularly with short passages or translated text, it maintains that the tool helps identify AI-generated content without compromising user privacy.
aiinsider