Apple reveals 'shocking evidence' from ex-employee's MacBook in OpenAI suit
Apple’s lawsuit against OpenAI over an ex-employee’s alleged theft of hardware design schematics is prompting broader questions about corporate ethics, trade secrets, and the use of AI tools at work. Commenters highlight how OpenAI appears unusually willing to benefit from misappropriated IP, contrast this with more “mature” firms that quickly distance themselves from tainted employees, and debate whether feeding confidential data into AI agents creates irreversibly “contaminated” models. The thread also surfaces practical concerns about employees using work devices and cloud accounts for personal or mixed use, given the legal and privacy risks when those devices become evidence.
Perceptions of OpenAI and Corporate Ethics
- Many see OpenAI’s alleged use of stolen Apple materials as desperate, unprofessional, and fitting a broader “move fast, ask forgiveness later” culture in tech.
- Some argue today’s tech leadership lacks prior-era ethical norms; others say this is not unique to tech.
- Comparisons are made to cases where competitors refused stolen IP (e.g., Coca‑Cola recipe) as a contrast to what OpenAI is accused of doing.
Employee IP Theft and Corporate Responses
- Multiple anecdotes: staff at various firms stealing databases, code, chemical libraries, or financial models; mature organizations typically fire and distance themselves quickly.
- Commenters emphasize that sophisticated firms (e.g., trading shops) have vast resources and aggressively pursue such theft.
- Some speculate OpenAI may have incentivized or at least failed to discourage this behavior; others think that’s unclear.
Trade Secrets, LLMs, and “Tainted” Models
- Apple’s claim: feeding trade secrets into an AI agent can create irreversible, propagating use of those secrets.
- This triggers debate:
- One side: this logic could undermine the “fair use” rationale for training on copyrighted works or GPL code.
- Others note trade secrets are legally distinct from copyright and the argument may not generalize.
- Clean-room analogies are discussed: use separate “dirty” and “clean” LLM runs or agents, similar to human reverse‑engineering practices, though trust in such separation is questioned.
- Some cite research showing large models can memorize entire books, bolstering concerns about embedded IP.
Human Knowledge vs AI Training
- Several distinguish humans taking experience to new employers (generally allowed, especially in California) from exporting company-trained AI agents, which look more like company property.
- There is concern that as skills are externalized into company-owned agents, workers become more “alienated” from their own expertise.
Work Devices, Privacy, and Discovery
- Strong consensus: don’t use work laptops/accounts for personal email or job hunting; everything may be monitored and discoverable in litigation.
- Apple’s iCloud setup blurring personal/work boundaries is seen as risky; some say employees are encouraged to “live on” work devices, others push back.
- Several note legal nuance: employer access vs. subpoenas, jurisdictional differences, and contracts that quietly waive privacy expectations.