Two weeks ago we walked through the CLAUDE rug pull factory — 15 fake Claude-branded contracts shipped on Ethereum in 14 days. That post focused on the latest wave (April 24 to May 7). What it did not cover is the older record: a single CLAUDE token deployed back in March that drained ~$314K from buyers across six days.
This is the receipt for that rug, plus the five others.
The scoreboard
Every Claude-branded ERC-20 with a confirmed liquidity pull on Uniswap v2 over the last 180 days. “Profit” = ETH withdrawn by the deployer minus ETH initially seeded into the pool. All six contracts are unverified on Etherscan, all six score 70+ on the RektRadar scale.
| Symbol | Name | Address | Profit | Rug date |
|---|---|---|---|---|
| Claude | Claude | 0x43baa6…1695 | +165.64 ETH | 2026-03-27 |
| CLAUDE | Claude | 0x0a2dc5…42d3 | +8.67 ETH | 2026-02-07 |
| 🙌PPC | Peppy Claude | 0x226f69…0615 | +1.19 ETH | 2026-05-07 |
| CORNER | Claude’s Corner | 0x19088c…076a | +0.99 ETH | 2026-02-25 |
| CLAUDSTR | Claude Strategy | 0x6165b7…da30 | +0.84 ETH | 2026-05-02 |
| KellyClaud | KellyClaude | 0x6260b3…1046 | ~0 ETH | 2026-02-22 |
Total: 177.33 ETH extracted — roughly $337K at 1900 ETH/USD. The top entry alone accounts for 93% of that.
$Claude (0x43baa6) — the $314K rug
Token deployed 2026-03-26. Pool 0x45e95dfc8b507de2401fc3fa73dfe5e8edfb04e2 on Uniswap v2. Risk score: 80/100. Flags: honeypot, low_liquidity, buy_failed, sell_failed, unverified_contract, unverified_bytecode_analyzed. The contract was never verified on Etherscan, which alone should have been enough to scare off careful buyers.
It was not.
Timeline
The deployer (0x6575e1...228E) seeded the pool with 30 ETH of liquidity at block 24,739,036. Then, over the next six days, drained it back out in stages:
| Date | Action | ETH | Block |
|---|---|---|---|
| 2026-03-26 03:26 | mint (LP add) | +30.00 | 24,739,036 |
| 2026-03-27 01:46 | burn | -28.77 | 24,745,703 |
| 2026-03-28 03:05 | burn | -29.78 | 24,753,257 |
| 2026-03-29 03:08 | burn | -29.54 | 24,760,456 |
| 2026-03-30 05:56 | burn | -29.56 | 24,768,468 |
| 2026-03-31 06:59 | burn | -28.82 | 24,775,950 |
| 2026-04-01 01:24 | burn | -49.17 | 24,781,450 |
Six burns. Each between 28 and 50 ETH. The math here only works because new buyers kept walking into the pool between drains — every burn empties the pool, and the next 24 hours buyers refill it via swaps. The deployer just kept coming back to siphon what landed in the pool overnight. Then on April 1, one final 49 ETH pull and the contract went silent.
The pattern is unusual. Most rugs we document are single-shot: deployer adds 30 ETH, lets buyers in for 6 hours, pulls everything in one transaction. This one ran for six days. The contract sat there on Etherscan, unverified, with a steady stream of victims, and the deployer drained the pool every morning around the same time UTC.
Why “Claude”
The token launched on March 26 — the exact week Anthropic’s Claude was top-of-funnel in Google searches across crypto Twitter. The brand-jack is the same playbook we walked through in the AI brandjack post and the recent CLAUDE factory case study — the AI lab itself never issued a token, so any ERC-20 named “Claude” on mainnet captures Google traffic for free.
What is different here is that 0x43baa6 got there first. The big mid-March wave of CLAUDE searches landed on a single contract with no competition. By the time copycats spun up in April, the original had already exited.
Five smaller rugs, same playbook
The other five rugs are textbook by comparison:
- CLAUDE (0x0a2dc5), +8.67 ETH on Feb 7. First Claude-branded rug we tracked, before the AI-brand meta really took off on Twitter.
- 🙌PPC Peppy Claude, +1.19 ETH on May 7. Last week’s, recent enough that some of you may have seen it on the trending lists.
- CORNER Claude’s Corner, +0.99 ETH on Feb 25. Sub-1 ETH but verified the same
creator_holds_all_lpflag. - CLAUDSTR Claude Strategy, +0.84 ETH on May 2. The “strategy” suffix is a recurring brand-jack vector — buyers who Google “Claude AI strategy” find it.
- KellyClaud KellyClaude, ~0 ETH on Feb 22. Failed rug — the buyer pool dried up before the deployer could extract more than dust.
None of the five was verified on Etherscan. None had a credible deployer history. All shared the same RektRadar flag fingerprint as the big one — just smaller buyer pools.
Detection in real time
The data above all comes from a single endpoint: /api/analytics/rug-metrics?since=180d&search=claude on the RektRadar app. Every CLAUDE-branded rug on Ethereum that ran a mint -> burn cycle on Uniswap v2 with a 90% drain ratio is in there. Searching “claude” also surfaces the five smaller ones in the same response.
Every contract above also has a per-token report at /result/<addr>. Click any of the addresses in the table to see flag breakdown, deployer wallet history, and the bytecode hash bucket the contract clusters with.
Three takeaways for buyers
- Anthropic does not have a token, period. Any ERC-20 named Claude on Ethereum mainnet is by definition not affiliated with the lab. If you remember nothing else, remember this for AI-brand tokens broadly: the AI lab itself is not the issuer.
- Unverified contract = walk away. All six rugs above were unverified on Etherscan. That alone is a 70+ score on our scale. Verified contracts are not safe by default, but unverified ones are unsafe by default.
- The first deployer wins SEO; later ones recycle the same victims.
0x43baa6ran for six days because it owned the search query “Claude crypto” before any competing contract existed. Every subsequent CLAUDE token competed for whatever scraps Google indexed next. If you find a Claude-branded token that has been trading for more than 24 hours, the rug has already happened — you are buying into the post-drain pool.
We will keep documenting these as they happen. The next AI brand will trend, the next factory will spin up, and the same playbook will repeat — only the symbol changes.
You can scan any CLAUDE address from the table above on the $CLAUDE scam tokens catalog on Ethereum. Free, fresh data, both buy and sell simulations — no signup required.