261–270 of 464 posts
Earlier quoted context omitted.
Turns out Google also hates 500, 174, 833, 285 and 302 (from generating files from -1000 to 1000).
Another bunch of numbers got flagged (186, 451, 336, 173, 266). I deleted the experiment, just in case I got my account deleted for too many naughty numbers.
I'm curious if it was related to the file name. I created a few 1-byte files with just "1" in them, with different names, including "output04.txt". No problems so far. Also uploaded variations with "\n" and "\r\n" after the "1". And enabled sharing to anyone with the link. No issues so far. Google drive does support metadata like a description and comments. I wonder if someone posted some copyrighted text in a commen…
I wonder if this is an incredible hash collision. hash("1") == hash(disney_movie.mp4)
I'm curious if it was related to the file name. I created a few 1-byte files with just "1" in them, with different names, including "output04.txt". No problems so far. Also uploaded variations with "\n" and "\r\n" after the "1". And enabled sharing to anyone with the link. No issues so far. Google drive does support metadata like a description and comments. I wonder if someone posted some copyrighted text in a commen…
Turns out Google also hates 500, 174, 833, 285 and 302 (from generating files from -1000 to 1000).
I'm curious if it was related to the file name. I created a few 1-byte files with just "1" in them, with different names, including "output04.txt". No problems so far. Also uploaded variations with "\n" and "\r\n" after the "1". And enabled sharing to anyone with the link. No issues so far. Google drive does support metadata like a description and comments. I wonder if someone posted some copyrighted text in a commen…
Update: It has now flagged the file with "0\n" in it. So it's official, Both 0 and 1 are copyrighted :) https://imgur.com/a/xMgh6Xn
Earlier quoted context omitted.
I definitely agree, but in america it's impossible to be legally employed without your employer having your SS# on record.
Well, yes. Employers need to talk to the Social Security Administration regarding your social security withholdings...
The same goes for credit card and bank account numbers. As anyone you've ever paid with your card or by check has access to these they can't be considered particularly secret. The problem is that the system barely has any authentication built in. 4-digit PINs and hastily handwritten signatures only an expert can verify do not offer reasonable security. Chip cards are a bit better (non-clonable, require physical possession) but only work for in-person transactions.
Earlier quoted context omitted.
Just tested this myself. Got the same results as you while using different file names. Difficult to believe that Google has become so internally dysfunctional that it hasn't fixed this technical embarrassment twenty hours after first being being seen, and then five hours after being reported on front page of Hacker News.
It only affects sharing so it isn't a disaster if this happens, probably P3 medium if you went by their bug bounty priorities.
"A review cannot be requested for this restriction" ML enforcing rules is bad enough, but not allowing false positives to be corrected is ridiculous. This is why I would never consider g-suite for any business application. Otoh, I think there is a legitimate business to be made helping small businesses and individuals secure themselves against arbitrary behavior from big tech. This kind of thing can have serious cons…
How long until cloud providers are forced to scan block devices, so you can't even self-host a file erroneously flagged by the ML? (Actually, how long until this is integrated with the "security engine" in CPUs, so you can't even hide with encryption? Everyone is talking about on-device voice-recognition and CSAM recognition, but no doubt the copyright lobby is going to push hard for this as well.)
(... I have actually used this in the past to work around my employer's moronic proxy)
Earlier quoted context omitted.
Turns out Google also hates 500, 174, 833, 285 and 302 (from generating files from -1000 to 1000).
This is relatively new then. As part of some testing last summer for how to handle a migration from one Drive to another, I generated a folder that had I think 10k files in it. They were all named 1.txt, 75.txt, 8492.txt and their contents matched the filename (i.e. 1, 75, 8492). They never got flagged for me. That was for a Google Workspace account, though, and they were deleted a few days later once I was done with…
Earlier quoted context omitted.
It'd be nice if the law had a escrow appeal process. Alleged violator now posts a $100 escrow, now accuser has to do the same. Then Google reviews it, makes a decision, and loser has to pay for it.
It should be the other way around. The claimant must "bet" $100 that he owns the copyright. Then the defendant can call the bluff and say "I raise $100000 that you do not in fact own this output04.txt file with a 1 in it". If the the claimant still thinks he can win he can call the $100k and prove ownership. Otherwise the defendant just made $100. How cool would that be?
E.g.: You can trust a legally enforced warranty with full refunds guaranteed by the government, because it costs real money to the manufacturer. You can't trust a "Best Quality!" sticker. It basically costs nothing. It's just words.
Copyright protection laws are the same kind of thing. While the marginal cost of enforcement is zero, there is similarly zero incentive to do it correctly and respectfully of the law.
If there was enforced financial penalties for each screw up, then it is assured that any errors like this will be ironed out very quickly.
No penalty? No bug fixing!