發生了什麼事
The Next Web reports that micro1 submitted a court filing offering $12.5 million for records from Spirit Aviation Holdings, which stopped flying in May and is being liquidated. The proposed purchase would exclude disciplinary, investigatory and union-related materials, and would store the data in the United States. Google’s competing agreement is scheduled for judicial consideration on September 9.
The Next Web reports that micro1, an AI training-data company, offered $12.5 million for Spirit Aviation Holdings’ internal records, exceeding Google’s agreed $10 million price by $2.5 million. The report attributes the existence of the offer to a court filing and says Bloomberg News reported on it. The source does not independently provide the filing or confirm the bid’s acceptance.
According to The Next Web, the records include about 500 million Microsoft Teams items, 100 million emails and roughly 16 million customer chat sessions. The report says Google’s agreement excluded customer chats, loyalty records and call recordings, while micro1’s offer names the customer sessions. Google says it would not receive personal information and would pay a third party to remove sensitive customer details.
The reported micro1 proposal would have an ombudsman selected by Spirit’s advisers rather than the buyer, would keep the data in the United States, and would exclude disciplinary and investigatory material as well as records tied to collective bargaining with Spirit’s unions. The unions have challenged the Google sale on privacy grounds, according to the source.
A judge is expected to consider Google’s purchase on September 9. The Next Web says courts rarely reopen auctions after they close, but the source does not establish whether the court has ruled on micro1’s offer or whether either transaction will proceed.
為什麼這很重要
The reported bid highlights the growing value of operational records as training data for AI systems, while exposing unresolved questions about consent, privacy, deidentification and secondary use. The records reportedly include hundreds of millions of workplace communications and customer interactions, making control over access and reuse consequential for employees, customers and any future model developer.
The proposed sale concerns unusually large archives of workplace and customer communications. If acquired for AI training, those records could influence future models or data products, but the source does not establish what model micro1 would train, whether the records can legally be used for that purpose, or whether any system has already been trained on them.
The dispute also shows why the label “deidentified” is not by itself proof that data cannot identify people. The Next Web discusses European legal standards concerning pseudonymized data and model training, but it also states that those rules do not govern Spirit’s American liquidation. The article’s legal analysis is not an independent legal determination.
For employees and customers, the practical issue is whether sensitive communications can be transferred, linked across systems or reused beyond the purposes for which they were originally collected. The source reports proposed safeguards and objections, but provides no independent audit, technical assessment or evidence of actual privacy protection.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
Which training-log entry describes one completed epoch?
接下來看什麼
The key next step is the court’s September 9 consideration of Google’s agreement and whether micro1’s higher offer can reopen the completed sale process. Watch for clarification about which records are ultimately included, how privacy protections are independently overseen, and whether Spirit’s unions or affected customers challenge the transaction further.
The court’s September 9 review should clarify whether Google’s agreement remains in force, whether micro1’s counteroffer is procedurally admissible, and which records are covered. The source does not say whether the judge will compare the offers on privacy terms, price, or both.
Further attention should go to the treatment of customer chat sessions, loyalty records, call recordings and referential links between pseudonymous records. The source does not confirm whether these materials will be transferred or how they would be technically separated.
The source does not document micro1’s access conditions, product, pricing or model-development plans. It also does not independently confirm Google’s deidentification process, the proposed ombudsman’s authority, or the outcome of the unions’ privacy challenges.