Existing whistleblower laws don't cover this. They should.
But to get there, we'll need to think beyond historical whistleblower protections.
Two big barriers we'll need to rethink:
1. Typically, reporting has to go to a government.
That's understandable generally -- policymakers want to incentivize coming to officials + companies worry about sharing issues with third parties.
But when private auditors are so central to your governance regime, that's not tenable.
If the government is relying on auditors to oversee companies, you can't have a system where employees fear for their jobs if they cooperate with auditors.
Once companies are required to cooperate with auditors by law, it seems a natural extension to require that they not interfere with the work of those auditors via retaliating against their employees.
--
2. Typically protection is limited to reporting law-breaking -- other valid grievance typically aren't enough.
That's a problem in AI, because there just aren't that many laws to violate + increasing risk is worth reporting before the BadThings TM happen.
Laws like SB 53 and the proposed federal WPA deal with this by also covering information the whistleblower has reasonable cause to believes poses a "specific and substantial danger to the public health or safety resulting from a catastrophic risk.
But here, the information may not actually pose a direct danger. It may, for example, just show gross incompetence and a disregard for safety. That's relevant information to get to an auditor even if it doesn't meet the general whistleblowing threshold.
Here too, I think all of this gets easier once auditors are required, and the right can be tied to ~interfering with a lawfully required investigation or audit.
But to get there, we'll need to think beyond historical whistleblower protections.
Two big barriers we'll need to rethink:
1. Typically, reporting has to go to a government.
That's understandable generally -- policymakers want to incentivize coming to officials + companies worry about sharing issues with third parties.
But when private auditors are so central to your governance regime, that's not tenable.
If the government is relying on auditors to oversee companies, you can't have a system where employees fear for their jobs if they cooperate with auditors.
Once companies are required to cooperate with auditors by law, it seems a natural extension to require that they not interfere with the work of those auditors via retaliating against their employees.
--
2. Typically protection is limited to reporting law-breaking -- other valid grievance typically aren't enough.
That's a problem in AI, because there just aren't that many laws to violate + increasing risk is worth reporting before the BadThings TM happen.
Laws like SB 53 and the proposed federal WPA deal with this by also covering information the whistleblower has reasonable cause to believes poses a "specific and substantial danger to the public health or safety resulting from a catastrophic risk.
But here, the information may not actually pose a direct danger. It may, for example, just show gross incompetence and a disregard for safety. That's relevant information to get to an auditor even if it doesn't meet the general whistleblowing threshold.
Here too, I think all of this gets easier once auditors are required, and the right can be tied to ~interfering with a lawfully required investigation or audit.
Kelsey Piper@KelseyTuoc · Oct 8This obviously makes it so that when OpenAI announces they're conducting an external investigation with regulators or auditors, everyone working with those regulators and auditors understands they'll be fired if the auditors or regulators learn too much.
Open quoted post → 2 6 0 33 1.6K 8
Tomek Korbak
Mikita Balesni
girish sastry
