> Then you wait and see if Amazon stops.

That’s if the requests are actually coming from Amazon, which seems very unlikely given some of the details in the post (rotating user agents, residential IPs, seemingly not interpreting robots.txt). The Amazon bot should come from known Amazon IP ranges and respect robots.txt. An Amazon engineer confirmed it in another comment: https://news.ycombinator.com/item?id=42751729

The blog post mentions things like changing user agent strings, ignoring robots.txt, and residential IP blocks. If the only thing that matches Amazon is the “AmazonBot” User Agent string but not the IP ranges or behavior then lighting your money on fire would be just as effective as hiring a lawyer to write a letter to Amazon.

Honestly, I figure that being on the front page of Hacker News like this is more than shame enough to get a human from the common sense department to read and respond to the email I sent politely asking them to stop scraping my git server. If I don't get a response by next Tuesday, I'm getting a lawyer to write a formal cease and desist letter.

I had to deal with some bot activities that used huge address space, and I tried something very similar, when condition confirming bot was detected I banned that IP for 24h

but due to amount of IPs involved this did not have any impact on about if traffic

my suggestion is to look very closely on headers that you receive (varnishlog in very nice of this and of you stare long enough at then you might stop something that all those requests have in common that would allow you to easily identify them (like very specific and usual combination of reported language and geo location, or the same outdated browser version, etc)

Uggh, web crawlers...

8ish years ago, at the shop I worked at we had a server taken down. It was an image server for vehicles. How did it go down? Well, the crawler in question somehow had access to vehicle image links we had due to our business. Unfortunately, the perfect storm of the image not actually existing (can't remember why, mighta been one of those weird cases where we did a re-inspection without issuing new inspection ID) resulted in them essentially DOSing our condition report image server. Worse, there was a bug in the error handler somehow, such that the server process restarted when this condition happened. This had the -additional- disadvantage of invalidating our 'for .NET 2.0, pretty dang decent' caching implementation...

It comes to mind because, I'm pretty sure we started doing some canary techniques just to be safe (Ironically, doing some simple ones were still cheaper than even adding a different web server.... yes we also fixed the caching issue... yes we also added a way to 'scream' if we got too many bad requests on that service.)

When I was writing a crawler for my search engine (now offline), I found almost no crawler library actually compliant with the real world. So I ended up going to a lot of effort to write one that complied with Amazon and Google's rather complicated nested robots files, including respecting the cool off periods as requested.

... And then found their own crawlers can't parse their own manifests.

> The consequence will almost universal blocks otherwise!

How? The difficulty of doing that is the problem, isn't it? (Otherwise we'd just be doing that already.)

What do you mean by "barely" respecting robots.txt? Wouldn't that be more binary? Are they respecting some directives and ignoring others?

Is there some way website can sell those Data to AI bot in a large zip file rather than being constantly DDoS?

Or they could at least have the curtesy to scrap during night time / off peak hours.

> The consequence will almost universal blocks otherwise!

Who cares? They've already scraped the content by then.

If they're AI bots it might be fun to feed them nonsense. Just send hack megabytes of "Bezos is a bozo" or something like that. Even more fun if you could cooperate with many other otherwise-unrelated websites, e.g. via time settings in a modified tarpit.

Global tarpit is the solution. It makes sense anyway even without taking AI crawlers into account. Back when I had to implement that, I went the semi manual route - parse the access log and any IP address averaging more than X hits a second on /api gets a -j TARPIT with iptables [1].

Not sure how to implement it in the cloud though, never had the need for that there yet.

[1] https://gist.github.com/flaviovs/103a0dbf62c67ff371ff75fc62f...

I work for Amazon, but not directly on web crawling.

Based on the internal information I have been able to gather, it is highly unlikely this is actually Amazon. Amazonbot is supposed to respect robots.txt and should always come from an Amazon-owned IP address (You can see verification steps here: https://developer.amazon.com/en/amazonbot).

I've forwarded this internally just in case there is some crazy internal team I'm not aware of pulling this stunt, but I would strongly suggest the author treats this traffic as malicious and lying about its user agent.

> The author is seeing requests from rotating residential IPs and changing user agent strings

This type of thing is commercially available as a service[1]. Hundreds of Millions of networks backdoored and used as crawlers/scrapers because of an included library somewhere -- and ostensibly legal because somewhere in some ToS they had some generic line that could plausibly be extended to using you as a patsy for quasi-legal activities.

[1] https://brightdata.com/proxy-types/residential-proxies

I wouldn't put it past any company these days doing crawling in an aggressive manner to use proxy networks.

> Whatever happened to courtesy in scraping?

Money happened. AI companies are financially incentivized to take as much data as possible, as quickly as possible, from anywhere they can get it, and for now they have so much cash to burn that they don't really need to be efficient about it.

> Whatever happened to courtesy in scraping?

When various companies got signal that at least for now they have a huge overton window of what is acceptable for AI to ingest, they are going to take all they can before regulation even tries to clamp down.

The bigger danger, is that one of these companies even (or, especially) one that claims to be 'Open', does so but gets to the point of being considered 'too big to fail' from an economic/natsec interest...

Mind sharing a decent robots.txt and/or user-agent list to block the AI crawlers?

The same thing that happened to courtesy in every other context: it only existed in contexts where there was no profit to be made in ignoring it. The instant that stopped being true, it was ejected.

> What if instead it was possible to feed the bots clearly damaging and harmfull content?

With all respect, you're completely misunderstanding the scope of AI companies' misbehaviour.

These scrapers already gleefully chow down on CSAM and all other likewise horrible things. OpenAI had some of their Kenyan data-tagging subcontractors quit on them over this. (2023, Time)

The current crop of AI firms do not care about data quality. Only quantity. The only thing you can do to harm them is to hand them 0 bytes.

You would go directly to jail for things even a tenth as bad as Sam Altman has authorized.

I've seen this tarpit recommended for this purpose. it creates endless nests of directories and endless garbage content, as the site is being crawled. The bot can spend hours on it.

https://zadzmo.org/code/nepenthes/

If with damaging content you mean incorrect content, another comment in this thread has a user doing what you said https://marcusb.org/hacks/quixotic.html

You should check your websites like grass dot io (I refuse to give them traffic).

They pay you for your bandwidth while they resell it to 3rd parties, which is why a lot of bot traffic looks like it comes from residential IPs.

How does it compare to common crawl?

Probably not copyright infringement. But it is probably (hopefully?) a violation of CFAA, both because it is effectively DDoSing you, and they are ignoring robots.txt.

Maybe worth contacting law enforcement?

Although it might not actually be Amazon.

On what legal basis?

I put my personal site behind Cloudflare last year specifically to combat AI bots. It's very effective, but I hate that the web has devolved to a state where using a service like Cloudflare is practically no longer optional.

If you go to a police station and ask them to arrest Amazon for accessing your website too often, will they arrest Amazon, or laugh at you?

While facetious in nature, my point is that people walking around in real brick and mortar locations simply do not care. If you want police to enforce laws, those are the kinds of people that need to care about your problem. Until that occurs, youll have to work around the problem.

Author of the article here. The behavior of the bot seems like this:

  while true {
    const page = await load_html_page(read_from_queue());
    save_somewhere(page);
    foreach link in page {
      enqueue(link);
    }
  }

Having everything behind auth defeats the point of making the repos public.

Depends on what you are hosting. I found that source code repository viewers in particular (OP mentions Gitea, but I have seen it with others as well) are really troublesome: Each and every commit that exists in your repository can potentially cause dozens if not hundres of new unique pages to exist (diff against previous version, diff against current version, show file history, show file blame, etc...). Plus many repo viewers of them take this information directly from the source repository without much caching involved, as it seems. This is different from typical blogging or forum software, which is often designed to be able to handle really huge websites and thus have strong caching support. So far, nobody expected source code viewers to be so popular that performance could be an issue, but with AI scrapers this is quickly changing.

Gitea in particular is a worst case for this. Gitea shows details about every file at every version and every commit if you click enough. The bots click every link. This fixed cost adds up when hundreds of IPs are at different levels of clicking of every link.

Seems some of these bots are behaving abusively on sites with lots of links (like git forges). I have some sites receiving 200 requests per day and some receiving 1 million requests per day from these AI bots, depending on the design of the site.

They said that it is coming from different ip addresses every time, so fail2ban wouldn't help.

I bet that URL (or a predecessor) resolves to a 404 page that has a broken link on it. technically your problem.

No. Feed them shit. Code with deliberate security vulns and so on.

Not a crawler writer but have FAFOd with data structures in the past to large career success.

...

The closest you could possibly do with any meaningful influence, is option C, with the general observations of:

1. You'd need to 'randomize' the generated output link

2. You'd also want to maximize cachability of the replayed content to minimize work.

3. Add layers of obfuscation on the frontend side, for instance a 'hidden link (maybe with some prompt fuckery if you are brave) inside the HTML with a random bad link on your normal pages.

4. Randomize parts of the honeypot link pattern. At some point someone monitoring logs/etc will see that it's a loop and blacklist the path.

5. Keep up at 4 and eventually they'll hopefully stop crawling.

---

On the lighter side...

1. do some combination of above but have all honeypot links contain the right words that an LLM will just nope out of for regulatory reasons.

That said, all above will do is minimize pain (except, perhaps ironically, the joke response which will more likely blacklist you but potentially get you on a list or a TLA visit)...

... Most pragmatically, I'd start by suggesting the best option is a combination of nonlinear rate limiting, both on the ramp-up and the ramp-down. That is, the faster requests come in, the more you increment their 'valueToCheckAgainstLimit`. The longer it's been since last request, the more you decrement.

Also pragmatically, if you can extend that to put together even semi-sloppy code to then scan when a request to a junk link that results in a ban immediately results in another IP trying to hit the same request... well ban that IP as soon as you see it, at least for a while.

With the right sort of lookup table, IP Bans can be fairly simple to handle on a software level, although the 'first-time' elbow grease can be a challenge.

The author's pronouns can be found here: https://github.com/Xe

It's a regular expression.

I think there’s a difference between crawling websites at a reasonable pace instead of just hammering the server to the point it’s unusable.

Nobody has problems with the Google Search indexer trying to crawl websites in a responsible way

This submission has nothing to do with IP laundering. The bot is straining their server and causing OP technical issues.

Personally I'm not trying to block the bots, I'm trying to avoid the bandwidth bill.

I've recently blocked everything that isn't offering a user agent. If it had only pulled text I probably wouldn't have cared, but it was pulling images as well (bot designers, take note - you can have orders of magnitude less impact if you skip the images).

For me personally, what's left isn't eating enough bandwidth for me to care, and I think any attempt to serve some bots is doomed to failure.

If I really, really hated chatbots (I don't), I'd look at approaches that poison the well.

HN isnt a monolith

Most of those of those artists aren’t any better though. I’m on a couple artists’ forums and outlets like Tumblr, and I saw firsthand the immediate, total 180 re: IP protection when genAI showed up. Overnight, everybody went from “copying isn’t theft, it leaves the original!” and other such mantras, to being die-hard IP maximalists. To say nothing of how they went from “anything can be art and it doesn’t matter what tools you’re using” to forming witch-hunt mobs against people suspected of using AI tooling. AI has made a hypocrite out of everybody.