OpenAI employees warned about operational security lapses for years, sources say
Former employees say broad internal access and weak reviews left OpenAI exposed to insider and outside intrusions as competition pushed speed over security.
- Before ChatGPT's late 2022 launch, current and former OpenAI employees warned leaders that lax security practices could be exploited by foreign adversaries; the company has recently shifted strategy, reassigning engineers and launching security reviews.
- Five employees reported that management prioritized rapid development over security, leaving sensitive code stored in a single 'monorepo' and unreviewed sandboxes vulnerable to attack, researcher Leopold Aschenbrenner warned in 2023.
- In August, senior executive Thibault Sottiaux reassigned one quarter of OpenAI's production engineers to improve security infrastructure, addressing long-standing employee concerns about open access to research on GitHub and Slack.
- The Justice Department charged former OpenAI engineer Abdulrahman Alfozan in October with planting a hidden recording device in an executive conference room, while researchers using Anthropic models breached multiple ChatGPT employee accounts.
- OpenAI spokesperson Elana Widmann stated the company "invests heavily in its security program," though former researcher Will DePue argues competitive pressure forces AI leaders to "move fast" rather than secure systems against adversaries.
8 Articles
8 Articles
OpenAI staff warned leaders of security vulnerabilities long before spy device discovery, say sources
Employees raised operational-security concerns for years, several souces say.They worried about openness of code, lack of security review, insider threats.OpenAI said it will keep strengthening defences as technology, threats evolve.
OpenAI employees warned about operational security lapses for years, sources say
When OpenAI disclosed in October that it found a recording device planted in an executive conference room, a number of former employees had the same reaction — they were not surprised.
According to ChainCatcher, according to foreign media reports, OpenAI employees warned management long before ChatGPT was launched that access to company code and research data was too broad, which could lead to sensitive technology leaks. Internal risks include too broad codebase permissions and extensive storage of research materials on GitHub and Slack. OpenAI has strengthened security construction and will launch a new round of special secur…
Workers reported fear of espionage, but revealed that leaders prioritized the development of data protection products
Coverage Details
Bias Distribution
- 75% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
