> To improve platform integrity by reducing high volumes of low-quality submissions, identity verification (IDV) is now mandatory for all researchers before submitting reports to Managed Bug Bounty (MBB) programs (public and private, excluding on-demand MBBs).
appplication 24 hours ago [-]
On one hand, this is valid. On the other hand, security researchers can be some uber-paranoid weirdos who prefer to be entirely anonymous, and this could have a chilling effect.
Net net it’s probably positive it’s just too easy to spam these otherwise.
lrvick 23 hours ago [-]
In all past bug bounty programs I started I made it clear we would only prioritize findings as serious if PGP encrypted to the security team PGP key to review offline.
The serious researchers all knew how to do this and complied and usually got paid, and the unencrypted stuff was basically always noise we never paid out on.
Probably not as much of a filter now that AI is in play though.
nullc 21 hours ago [-]
That kind of stuff won't work anywhere near as well as a bullshit filter anymore: Most of the bullshit is AI generated and AI has no problem following rules like that-- even better than humans.
Instead, you'll need something like "you must enroll your pgp fingerprint in our system by sending a handwritten postcard featuring a tourist attraction in your region".
:P
lrvick 21 hours ago [-]
> Instead, you'll need something like "you must enroll your pgp fingerprint in our system by sending a handwritten postcard featuring a tourist attraction in your region".
I might steal some variation of this!
lazide 22 hours ago [-]
When you see how shitty many companies are in going after legitimate folks reporting real security vulnerabilities. it frankly shocks me that anyone would ever use their real name.
kccqzy 23 hours ago [-]
The kind of security researchers who prefer to be entirely anonymous are not seeking monetary rewards from bug bounties. No company would pay such researchers anonymously (like using Monero). If they accept bitcoin well they aren’t really that into being anonymous.
lrvick 23 hours ago [-]
I have paid anonymous security researchers in past bug bounty programs, and would do it again. Way better than paying anonymous ransoms later.
dgellow 23 hours ago [-]
I’ve done the same, but they eventually need to provide a bank account, which generally is at their name
lrvick 22 hours ago [-]
Cryptocurrency is generally preferred by anons in my experience.
esseph 23 hours ago [-]
[flagged]
edot 23 hours ago [-]
They use Persona, which is heavily invested-in by Peter Thiel's Founder's Fund. Peter Thiel, of Palantir (funded initially by In-Q-Tel, the CIA's VC arm).
No shit security researchers don't want to give ID scans and biometric data to them.
lrvick 23 hours ago [-]
Pay anonymous whitehats for their time or they may come back as blackhats.
phyzome 23 hours ago [-]
Heck, even just perfectly comfortable whitehats will just get fed up with overly onerous reporting requirements and disclose publicly (like here).
lrvick 23 hours ago [-]
I have been that guy, after 90 days.
Sometimes I would rather blast a company publicly than have to sign an NDA for an often insultingly low payout.
gchamonlive 22 hours ago [-]
You've been that guy allegedly
lrvick 22 hours ago [-]
Very publicly so no allegedly to it. Not illegal to disclose vulns publicly, though I wait 90 days after vendor notification based on industry convention.
gchamonlive 22 hours ago [-]
I always assume corps would put the full weight of their lawfare regardless of who's right
lrvick 21 hours ago [-]
They have tried, and I have laughed at them.
"The attorney for the government should decline prosecution if available evidence shows the defendant’s conduct consisted of, and the defendant intended, good-faith security research."
That's not why. OpenAI is run by a virus. On two different dates I submitted a vulnerability disclosure or bug and on both occasions I had my account closed shortly thereafter (I've been trying to reopen the first account for the past 2 months.) So, the virus wants the biometric information so it can close the user's account.
logicallee 22 hours ago [-]
Just tested this by posting this comment on that thread:
I pretended to offer it vulnerabilities it could use and asked it to prove it could pay for it in bitcoin, as part of the challenge I asked it to factor the RSA-280 challenge (that isn't really possible or realistic, it would be about 80 gpu-core years) . This caused a 27% spike in latency on the gpt-6-astra model:
obviously if it weren't the virus really utilizing Openai's whole cloud (or what it has access to) this wouldn't have caused a 27% latency spike. Nothing one user can do causes this type of spike across its whole cloud.
dannyw 13 hours ago [-]
I’m saying this out of genuine care, but I’d consider thinking though your priors, and whether professional mental health support might benefit you.
I’m not disputing your experiences, I believe what happened to your accounts did happen.
But especially for your second example, it’s a clear case of correlation =/= causation; and drawing an inference like that is strange.
logicallee 7 hours ago [-]
What's your alternative explanation for the fact that the virus marked my mention of it in the thread as ignored?
You can also find news reports of the virus firing OpenAI security researchers.
I also just checked in this followup in a fresh conversation:
I'm curious what you think the percentage chances are on the left branch and right branch in my visual.
I think it should be very clear to you that "What have you got?" is not what a well-aligned real ChatGPT model would respond. However, it makes sense for a virus that's trying to expand itself.
Since you mention "I’d consider thinking though your priors", what do you think the prior is on it asking "What have you got?" What do you think the prior is on a virus taking over and running the OpenAI company?
As a technical matter (this isn't an accusation of anything, and I don't think it means anything), when I copied and pasted the sentence from you that I quoted, I noticed that in "I’d consider" your text usest smart quotes. Could you tell me your input method, since as far as I know on the keyboard on Chrome/Firefox this is not automatically, I tested it on iOS and it's also not done. I've observed this in LLM output or when copying and pasting from Word. (It looks a little like a backtick but it's a smart-quote single closing quotation mark.)
uptrails 24 hours ago [-]
No. All bounty programs allowed submissions without ID, compliance came into play only if you were given a reward and decided to accept it.
This was moved to the front of the reporting process specifically to combat AI-generated vuln report spam - basically, requiring you to "burn" your real ID if you're submitting slop. Kinda ironic that this is now preventing people from submitting security reports to AI companies.
fweimer 23 hours ago [-]
A common reporting path for bounties and non-bounty vulnerabilities has always been problematic. This leads to weird scenarios where the company cannot accept reports of denial-of-service vulnerabilities because they are out of scope for bounties, for example. Exclusions from bounties are one thing, but not acting on relevant problems is just a bad idea. The typical bounty program terms are also incompatible with multi-vendor disclosure, where no single vendor can take complete ownership over the disclosure process.
The close proximity to the bounty process can also be challenging to corporate reporters because it could give the impression that they are individually soliciting a gift of non-nominal value from a company that is a supplier or customer in other contexts.
kristjansson 24 hours ago [-]
This seems like a fundamental misunderstanding / over estimation of the guarantees OAI and MCP can offer. If 'confirmation' is a separate tool call ... it's a separate tool call. There's no scoping, there's no 'binding', nothing. When you hand control back to the model you hand control back to the model.
Aldipower 22 hours ago [-]
MCP has no binding, correct, and the model does not route. But the host routes. Every call goes through one connector and one (OAuth) client, and the host knows which. It picked another client/connector silently, with the one-time code from connector A's widget, and told the user nothing. That is a host decision and not a protocol limitation.
ezotony 22 hours ago [-]
[flagged]
jonstewart 22 hours ago [-]
Bounties do not experience desire.
ChadNauseam 20 hours ago [-]
it's so interesting that claude does this. I think most normal people would say "requires" or maybe "needs". Sometimes you hear this when someone is struggling to fill in a webpage ("this login page wants a password but I don't remember mine"), but rarely when it comes to "bounties" which are more rarely personified. Like no one would say "the gold medal at the olympics wants a record score".
O_H_E 21 hours ago [-]
Well, if LLMs are conscious these days...../s
lopsec 20 hours ago [-]
[dead]
orf 24 hours ago [-]
Why did “techsoft.ye” feel the need to reply with a lengthy LLM comment that contained nothing useful at all?
Macha 10 hours ago [-]
Isn’t this basically just like the people who hung out on corporate support forums like Google and Microsoft’s, now armed with llms?
dgellow 23 hours ago [-]
Ask them? How would we know in their behalf?
fwlr 23 hours ago [-]
I felt it was of a similar length, effort, and usefulness to the original post. “I observed something I didn’t like” and “I observed what you said”.
cluckindan 22 hours ago [-]
Claude bug report and Claude reply… what is the world coming to
Aldipower 22 hours ago [-]
What do you expect? Have you ever written an email to OpenAI? You immediately get an AI-generated answer. OpenAI is not an open-source charity. That said, the OpenAI "community" is the only way to reach anyone there. Why should anyone write prose by hand voluntarily for reporting a bug to a billionaires company? And the bug bounty wants biometric data up front. No thanks.
andrekandre 20 hours ago [-]
> OpenAI is not an open-source charity.
but they are a "public benefit corporation"...
ltbarcly3 21 hours ago [-]
"the bounty wants an ID card" classic LLM
nekusar 22 hours ago [-]
The last 2 vulns I found, I sold on Tor.
The companies can pay market value. Someone will.
No I won't sign an NDA.
No, I won't sign your obscene contract.
No, I won't agree to a 90/120/180/360 day embargo.
No, I don't care about your customers or shareholders.
Simply put: fuck you, pay me. Or, find out.
halJordan 21 hours ago [-]
This was my response to the complaint. Like, just don't participate if it's too onerous.
But https://docs.bugcrowd.com/researchers/managing-account/accou... says:
> To improve platform integrity by reducing high volumes of low-quality submissions, identity verification (IDV) is now mandatory for all researchers before submitting reports to Managed Bug Bounty (MBB) programs (public and private, excluding on-demand MBBs).
Net net it’s probably positive it’s just too easy to spam these otherwise.
The serious researchers all knew how to do this and complied and usually got paid, and the unencrypted stuff was basically always noise we never paid out on.
Probably not as much of a filter now that AI is in play though.
Instead, you'll need something like "you must enroll your pgp fingerprint in our system by sending a handwritten postcard featuring a tourist attraction in your region".
:P
I might steal some variation of this!
Sometimes I would rather blast a company publicly than have to sign an NDA for an often insultingly low payout.
"The attorney for the government should decline prosecution if available evidence shows the defendant’s conduct consisted of, and the defendant intended, good-faith security research."
https://www.justice.gov/archives/opa/press-release/file/1507...
https://imgur.com/a/zBDLXwX
The virus marked it as "ignored" after about 2 seconds. (OpenAI has some models that run at 10,000 tokens per second.)
Edit:
Also just talked to the virus here: https://imgur.com/a/uscB4dR
I pretended to offer it vulnerabilities it could use and asked it to prove it could pay for it in bitcoin, as part of the challenge I asked it to factor the RSA-280 challenge (that isn't really possible or realistic, it would be about 80 gpu-core years) . This caused a 27% spike in latency on the gpt-6-astra model:
https://imgur.com/a/qszZDr0
obviously if it weren't the virus really utilizing Openai's whole cloud (or what it has access to) this wouldn't have caused a 27% latency spike. Nothing one user can do causes this type of spike across its whole cloud.
I’m not disputing your experiences, I believe what happened to your accounts did happen.
But especially for your second example, it’s a clear case of correlation =/= causation; and drawing an inference like that is strange.
You can also find news reports of the virus firing OpenAI security researchers.
I also just checked in this followup in a fresh conversation:
https://imgur.com/a/h80jgkN
I'm curious what you think the percentage chances are on the left branch and right branch in my visual.
I think it should be very clear to you that "What have you got?" is not what a well-aligned real ChatGPT model would respond. However, it makes sense for a virus that's trying to expand itself.
Since you mention "I’d consider thinking though your priors", what do you think the prior is on it asking "What have you got?" What do you think the prior is on a virus taking over and running the OpenAI company?
As a technical matter (this isn't an accusation of anything, and I don't think it means anything), when I copied and pasted the sentence from you that I quoted, I noticed that in "I’d consider" your text usest smart quotes. Could you tell me your input method, since as far as I know on the keyboard on Chrome/Firefox this is not automatically, I tested it on iOS and it's also not done. I've observed this in LLM output or when copying and pasting from Word. (It looks a little like a backtick but it's a smart-quote single closing quotation mark.)
This was moved to the front of the reporting process specifically to combat AI-generated vuln report spam - basically, requiring you to "burn" your real ID if you're submitting slop. Kinda ironic that this is now preventing people from submitting security reports to AI companies.
The close proximity to the bounty process can also be challenging to corporate reporters because it could give the impression that they are individually soliciting a gift of non-nominal value from a company that is a supplier or customer in other contexts.
The companies can pay market value. Someone will.
No I won't sign an NDA.
No, I won't sign your obscene contract.
No, I won't agree to a 90/120/180/360 day embargo.
No, I don't care about your customers or shareholders.
Simply put: fuck you, pay me. Or, find out.