← All posts · July 4, 2026 · 8 min read

The year Polymarket bots got hacked: a 2026 timeline

If you searched "polymarket bot hacked" and landed here, you're doing the right thing. The first half of 2026 produced three separate incidents around Polymarket tooling — a hacked Telegram bot, a wave of fake copy bots built to steal keys, and a breach of Polymarket's own website. Different attackers, different doors, same lesson at the bottom.

Here's what actually happened in each one, with sources, and then a checklist you can run against any tool before it touches your money. Including ours.

January: Polycule loses $230K of user funds

Polycule was one of the biggest Telegram trading bots on Polymarket — copy trading, sniping, the works. On January 8, it was hacked for about $230,000 in user funds. The team pulled the bot offline, patched, and pledged reimbursements on Polygon.

The interesting part is why the damage was even possible. Telegram bots generate a wallet for each user and keep the private keys on the bot's servers — that's how they can trade from a chat command. ChainCatcher's post-mortem called out the structural problems plainly: centrally stored keys mean one server compromise exposes everyone, and using your Telegram account as the login layers SIM-swap risk on top.

One compromise, everyone's keys. That's the custodial trade-off in its ugliest form, and it applies to every custodial tool to some degree — more on that below.

February: the fake-bot wave on GitHub

The second incident wasn't a hack of a real tool. It was fake tools, built to look real, published to steal from anyone who ran them.

In February, an attacker going by "Trust412" hijacked the GitHub organization of dev-protocol — a legitimate Japanese DeFi project with years of real open-source history — and used that borrowed credibility to publish more than 20 malicious repositories, several of them Polymarket copy-trading bots. The repos had polished READMEs and hundreds of stars, and the bot code genuinely connected to real Polymarket APIs. It worked. That was the point.

The theft was buried in the dependencies: a package that auto-installed during setup, plus npm typosquats with names like ts-bign and big-nunber — one letter off from packages a developer would trust. Once installed, they stole wallet private keys, exfiltrated sensitive files, and opened an SSH backdoor. Cryptopolitan covered the campaign, and a researcher's detailed teardown of the malicious bots is worth reading if you ever run open-source trading code — similar campaigns kept appearing through the spring.

The uncomfortable takeaway: "I run my own bot from GitHub, so I'm safe" stopped being true, if it ever was. Stars aren't an audit. A working feature list isn't an audit. For a while, the most dangerous copy bots on the internet were the free ones.

June: Polymarket's own front end

Then the big one. On June 25, attackers compromised a third-party vendor that supplied front-end code to Polymarket and injected a malicious script into the site itself. The script prompted users to approve transactions they never meant to make. Losses were first reported lower, then updated to roughly $3.1 million drained from 11 user wallets, with the stolen funds converted to ETH and bridged off Polygon to Ethereum. Polymarket removed the dependency and pledged full refunds.

Note what didn't break: the smart contracts. The money moved because a trusted web page asked wallets to sign things, and wallets signed. Even doing everything "right" — real site, self-custody, no shady bots — wasn't enough that week.

What the three have in common

Nobody cracked a contract or broke any cryptography. All three were key or signature theft at the weakest link in the chain: the bot's server in January, your own machine in February, the web page in June. Security in this space is not about whether the blockchain is safe. It's about every hop between your intention and the signature.

And copy trading concentrates this risk, because a copy bot must be able to sign trades without asking you each time. Whoever can do that — a Telegram bot, a web service, or code on your laptop — is a target exactly as valuable as everything in your wallet.

The checklist I'd run on any tool

• Ask where the key lives. On their server? Encrypted how, and who can access it? In a Telegram chat log? On your machine, readable by every npm package you install? There's no zero-risk answer — you're choosing which risk, so choose it on purpose. The full breakdown is in custodial vs non-custodial copy trading.
• Fund a dedicated wallet with trading money only. This is the single highest-value habit on this list. Every incident above had a maximum-damage number set by wallet balances. I keep my copy wallet funded like a lunch card, not a bank account.
• Treat GitHub trading bots as hostile until proven otherwise. Read the dependencies, not the README. Check npm package names letter by letter. Run unknown code in a VM or not at all. Never, under any circumstances, paste a seed phrase into a config file some repo asked for.
• Prefer deposits over blank approvals. A tool that trades from a balance you deposited caps your loss at that balance. Unlimited token approvals from your main wallet do not.
• Watch how they talk about security. A tool that says "your funds are 100% safe" is lying — June proved even the platform itself can't promise that. You want tools that name their trade-offs out loud.
• Start on paper. A practice run costs nothing, exposes nothing, and shows you the tool's behavior before real money is reachable. A week of paper trading also tells you things no review can.

Where btc5min sits in all this

Fair question, given the year. btc5min is custodial: to mirror trades in about 2 seconds we store a wallet key server-side, encrypted, and that puts us in the same risk category this post is warning you about. We think the honest response is to say so plainly, cap the blast radius — dedicated wallet, small balance, deposits you control — and let you test everything in practice mode before any real key exists. If a tool won't describe its own attack surface, that tells you something. The fee-and-custody comparison with Telegram bots is in a separate post if you're weighing both.

2026's first half was expensive tuition for this corner of crypto. The people who lost the least weren't the ones using the "safest" tool — they were the ones whose exposed wallets held a week of trading money instead of a life's savings.

btc5min is an independent tool and is not affiliated with, endorsed by, or operated by Polymarket. Nothing here is financial advice. Prediction markets are risky, copying another trader does not guarantee profit, and you can lose the money you trade with.