3 weeks ago
AI Threat to Corporate Trade Secrets as Knowledge Walks Out
Companies have special secrets that help them build their products, kind of like a secret recipe.
In the past, if a worker left to join another company, they might take papers or computer files that held those secrets.
This article talks about a big disagreement between Apple and OpenAI.
Apple says a former worker took secret information before joining OpenAI, and OpenAI says that is not true.
Long ago, when people took secrets, companies could prove it because there were documents and computer records to show.
For example, a car company called General Motors said a worker took secrets to another car company, and that company had to pay a huge amount of money.
But now, very clever computer programs called AI are getting better at helping people rebuild things from memory.
So a worker may not need to take any files — they can simply remember how things worked and use AI to recreate them.
This makes it much harder for companies to protect their secrets, and the old rules and laws are not ready for this new problem.
Apple alleges a former employee took confidential information before joining OpenAI, a claim OpenAI disputes and the courts will decide.
Frontier AI models such as Anthropic's Claude Fable 5, OpenAI's ChatGPT Sol, and Moonshot AI's Kimi K3 can compress coding, reasoning, and synthesis into one interface.
In 1997, Volkswagen agreed to pay General Motors $100 million and buy at least $1 billion in parts over the López de Arriortúa dispute.
TSMC settled with SMIC in 2005, with SMIC paying $175 million over six years plus cross-licensing and dismissal of pending legal actions.
The Waymo-Uber dispute over Anthony Levandowski's alleged download of more than 14,000 confidential files settled in 2018 for about $245 million in Uber equity.
- Who
- Apple and OpenAI are in a legal dispute; Apple alleges a former employee took confidential information before joining OpenAI, while OpenAI disputes the allegations.
- What
- A legal battle over alleged trade secret misappropriation, framed as a sign that frontier AI is making institutional knowledge easier for departing employees to recreate without leaving any trail.
- Where
- Described as a Silicon Valley courtroom drama, with earlier disputes involving companies including General Motors, Volkswagen, Renault, TSMC, SMIC, Waymo, and Uber.
- When
- The Apple-OpenAI dispute is current and before the courts; the article also cites prior cases from 1997, 2005, 2011, and 2018.
- Why
- Because advancing AI lowers the cost of reconstructing corporate knowledge from memory, weakening trade secret protections that historically depended on proof of transfer.
Traditional trade secret protection
AI-amplified brain drain
Trade secret misappropriation allegations
Traditional trade secret protection
Apple claims a former employee walked away with confidential information before joining OpenAI.
AI-amplified brain drain
OpenAI disputes those allegations, and the courts will ultimately decide the facts.
How to protect institutional knowledge
Traditional trade secret protection
Organisations should respond by tightening NDAs, expanding garden leave, monitoring exits more aggressively, and demanding heavier compliance rituals.
AI-amplified brain drain
The article argues those measures target an outdated threat model, since the most valuable corporate knowledge is now easier to reconstruct than to contain.
Key facts
- Core dispute
- Apple alleges a former employee walked away with confidential information before joining OpenAI; OpenAI disputes the allegations
- AI models cited
- Anthropic's Claude Fable 5, OpenAI's ChatGPT Sol, Moonshot AI's Kimi K3
- GM-Volkswagen settlement (1997)
- Volkswagen paid GM $100 million and agreed to buy at least $1 billion in parts
- TSMC-SMIC settlement (2005)
- SMIC to pay $175 million over six years, with cross-licensing and dismissal of legal actions
- Waymo-Uber settlement (2018)
- About $245 million in Uber equity
- Waymo allegations
- Anthony Levandowski allegedly downloaded more than 14,000 confidential files from Google's self-driving project
- Renault case (2011)
- Three senior executives suspended over suspected EV-programme leaks; case later found unfounded and Renault apologised
- Author
- Sreejith Sreedharan, technology analyst and author of 'Future of Work: AI Augmented Autonomous Decentralised'



