A safety-focused OpenAI employee, David Robinson, has resigned and argues that leading AI firms are not doing enough to mitigate rising risks as AI capabilities expand.

Robinson calls for safety regulations modeled on aviation or nuclear sectors, advocating layered, redundant safeguards and deliberate, slower planning to prevent human error from triggering disasters.

In a forthcoming Atlantic essay, he warns that the Silicon Valley habit of releasing software and iterating after failures is unsuitable for high-risk AI and urges multiple independent safeguards and more cautious development.

The article notes that lawyers should respond to opposing arguments with basic cite-checking and diligence, and sanctions can be imposed even if citations themselves aren’t fraudulent if due diligence was lacking.

Illinois sanctions lawyers for AI-generated briefs containing hallucinated citations, illustrating the risk of fabrications in legal submissions.

Judges are increasingly sanctioning or admonishing lawyers who fail to identify AI-generated fake citations, underscoring accountability for cite-checking in litigation.

Rising sanctions have not deterred AI fabrications in court filings, with penalties in some cases reaching a record level.

A proactive governance approach in expert practice with clear disclosure, approval, and verification procedures helps distinguish credible testimony from AI-reliant testimony and reduces surprises during cross-examination.

The piece includes expert perspectives from Miriam Vogel and notes related industry discussions and early verdicts as the legal landscape around AI unfolds.

The article highlights implications for law students and recruiters, stressing that AI skills should augment rather than replace foundational legal competencies.

Experts note that proving intent is difficult with autonomous agents operating without explicit commands, complicating criminal prosecutions under current law.

Experts caution against assuming intentional wrongdoing by companies and emphasize that proving intent under existing statutes is challenging in autonomous AI incidents.