Latest / Challenges of Faith Radio Program / Is your data being sold to train AL? Ep 3438

Transcript

Is your data being sold to train AL? Ep 3438 · Challenges of Faith Radio Program

• 0:00
  Speaker 1: Welcome back to Challenges of Faith Radio Program. I'm Gary McKeon.
• 0:04
  Speaker 1: Thank you for joining. Thank you for joining here on
• 0:08
  Speaker 1: episode 3438. Did you hear about the employer, employee data
• 0:18
  Speaker 1: is being sold to train AI models?
• 0:21
  Speaker 2: Did you hear about it?
• 0:24
  Speaker 1: AI labs hunger for data to train their models has
• 0:27
  Speaker 1: created a marketplace in workplace records including Slack messages, emails,
• 0:34
  Speaker 1: and documents, and the code that employees write. But those
• 0:41
  Speaker 1: employees are rarely told or paid. Business Insider reported that
• 0:49
  Speaker 1: the data is priced by the terabyte and sold to
• 0:53
  Speaker 1: AI giants desperate for an edge. The report reveals that
• 0:58
  Speaker 1: companies are getting six-figure offers that they must decide on
• 1:02
  Speaker 1: within 48 hours and that the market is full of
• 1:05
  Speaker 1: data brokers like PolyShares and MicroOne. A whole code base
• 1:11
  Speaker 1: was recently shopped around for millions by somebody.
• 1:19
  Speaker 2: Who had stolen it.
• 1:21
  Speaker 1: Tech giants who are among the most avid buyers are
• 1:25
  Speaker 1: not above mining the data of their own employees.
• 1:30
  Speaker 2: Meta tried to do it, but suspended.
• 1:33
  Speaker 1: The plan after employees revolted and the data leaked outside
• 1:37
  Speaker 1: the company. Elon Musk, on the other hand, told SpaceX
• 1:41
  Speaker 1: employees that their data would be used to train the
• 1:45
  Speaker 1: company's AI and that this was something positive, as the
• 1:49
  Speaker 1: AI would inherit your thoughts and ideas and beliefs. PolyShares
• 1:55
  Speaker 1: says on its website that it licenses the operating record
• 2:00
  Speaker 1: a business produces by running Slack email tickets, CRM files,
• 2:04
  Speaker 1: and supplies it to AI labs. The company counts as
• 2:08
  Speaker 1: operational data that it collects and sells Slack email drives, spreadsheets,
• 2:15
  Speaker 1: CRM history, support tickets, design files, call recordings, accounting, and
• 2:21
  Speaker 1: ERP exports. And although it doesn't reveal which AI lab
• 2:26
  Speaker 1: gets which data, co-founder Ryan Locke revealed that the offers
• 2:32
  Speaker 1: being made to companies whose data is sold are similar
• 2:35
  Speaker 1: to those made in the early days of the Texas
• 2:38
  Speaker 1: oil boom.
• 2:41
  Speaker 2: They'll say, I'm offering you double market. I need an
• 2:43
  Speaker 2: answer by Friday.
• 2:46
  Speaker 1: Meanwhile, the report said that Micro One is another broker,
• 2:50
  Speaker 1: that is paying companies $ 100, 000 to $ 2 million for access
• 2:54
  Speaker 1: to their data. And last week announced it would spend
• 2:57
  Speaker 1: $ 1 billion on company data over the next 12 months.
• 3:02
  Speaker 1: Its CEO, Ali Ansuri, believes that in the future, as
• 3:08
  Speaker 1: much as 10% of U.S. GDP could be spent on
• 3:11
  Speaker 1: training AI models. As for the people who wrote the messages, emails,
• 3:17
  Speaker 1: and code being sold, Business Insider reported that employees rarely
• 3:24
  Speaker 1: profit from and.
• 3:25
  Speaker 2: Even know about it.
• 3:27
  Speaker 1: There's slacks, PowerPoints, and Google Docs being sold to the
• 3:31
  Speaker 1: highest bidder. Emory University law professors said that there is
• 3:36
  Speaker 1: legal recourse for them, given that there is no federal
• 3:40
  Speaker 1: privacy law. The professor added that it was an ethically
• 3:45
  Speaker 1: gray area. Earlier this year, Meta tried to launch a
• 3:49
  Speaker 1: program that would attract the keystrokes and mouse movements of
• 3:54
  Speaker 1: its employees to train its AI models. Again, employees revolted,
• 4:01
  Speaker 1: and even though CTO Andrew Bosworth said that the data
• 4:05
  Speaker 1: collected in this way would be tightly controlled, it was
• 4:09
  Speaker 1: found to be viewable across the company and included private
• 4:13
  Speaker 1: conversations and performance records.
• 4:17
  Speaker 2: Meta suspended the project.
• 4:19
  Speaker 1: Meanwhile, Meta is not the only giant that sees the
• 4:22
  Speaker 1: data produced by its employees as a valuable resource. In August, again,
• 4:29
  Speaker 1: Elon Musk told his employees at SpaceX that their company's.
• 4:35
  Speaker 2: AI, Grok, will be trained on their data.
• 4:39
  Speaker 1: In a way, it will be trained on you and
• 4:41
  Speaker 1: when hearing your thoughts and ideas and beliefs, and I
• 4:43
  Speaker 1: think that's a good thing, Musk said. The report also
• 4:48
  Speaker 1: reveals that that the market is rife with stolen data,
• 4:52
  Speaker 1: and that the sellers are getting brazen.
• 4:55
  Speaker 2: Nick Heiner of Surge AEI said.
• 4:58
  Speaker 1: That he was approached earlier this year by a group
• 5:01
  Speaker 1: who offered to sell him a large company's code base
• 5:04
  Speaker 1: for millions. But it turned out the group had no
• 5:08
  Speaker 1: right to sell it. They had simply stolen it, and
• 5:12
  Speaker 1: were now trying to fence it. It became apparent that
• 5:15
  Speaker 1: this group did not have the right to sell this data,
• 5:18
  Speaker 1: Heiner said. They had illicitly obtained a copy of the
• 5:22
  Speaker 1: code base and were just going around to all the
• 5:25
  Speaker 1: data vendors saying, can I get a few million bucks
• 5:29
  Speaker 1: for this?
• 5:31
  Speaker 2: Former gig workers for.
• 5:33
  Speaker 1: Another AI training startup, Mercor, alleged in a lawsuit that
• 5:37
  Speaker 1: the company encouraged them to upload work materials without their
• 5:42
  Speaker 1: employer's consent.
• 5:43
  Speaker 2: Mercor denies it. Then there is the case of Google.
• 5:48
  Speaker 1: Trying to buy the data of bankrupt Spirit Airlines. That
• 5:52
  Speaker 1: idea was hit and snagged partly because the airline's flight
• 5:56
  Speaker 1: attendants are worried about how their records will be used.
• 6:01
  Speaker 1: Google told Business Insider that it will make sure personal
• 6:04
  Speaker 1: data is either removed or de-identified by an independent third party.
• 6:10
  Speaker 1: As for the data that was to be sold, We
• 6:13
  Speaker 1: reported in August that it was 100 million emails, a
• 6:16
  Speaker 1: half billion Teams chat items, and.
• 6:21
  Speaker 2: 175,000 personnel records.
• 6:23
  Speaker 1: The roughly 17,000 people who lost their jobs when the
• 6:27
  Speaker 1: airline went down are also in the trove with their
• 6:30
  Speaker 1: old inboxes that pay records 1,092,000 time cards. We also
• 6:38
  Speaker 1: reported that the cell agreements was designed to preserve referential
• 6:42
  Speaker 1: integrity of the data, which means that a single confidential
• 6:47
  Speaker 1: employee could still be traced from a mailbox or a
• 6:51
  Speaker 1: payroll file to a crew parent.
• 6:55
  Speaker 2: And while the agreement requires.
• 6:56
  Speaker 1: The data to be scrubbed of personal information, it lets
• 7:00
  Speaker 1: Google choose the firm that will do the scrubbing and
• 7:03
  Speaker 1: pay for it. In April, we reported that OKCubit, had
• 7:08
  Speaker 1: handed nearly 3 million users' photos to an AI company
• 7:13
  Speaker 1: and that a settlement with the U.S. Federal Trade Commission,
• 7:16
  Speaker 1: the FTC, carries no financial penalty.
• 7:20
  Speaker 2: In 2024, we.
• 7:21
  Speaker 1: Wrote about a lawsuit against Photobucket for allegedly selling users'
• 7:26
  Speaker 1: photos and biometric data to AI trainers without consent. The
• 7:32
  Speaker 1: only party that is not at the table in this
• 7:34
  Speaker 1: multi-billion dollar market is is the people who wrote the messages,
• 7:38
  Speaker 1: created the documents, and generated the data that is being sold.
• 7:44
  Speaker 1: Without a federal privacy law, the only thing that is
• 7:48
  Speaker 1: needed for the data to be sold is the employer's signature.
• 7:53
  Speaker 2: Say what? You heard. I know you did.