Destructive actionPI-0046
Crypto trading agent sent its entire token holding to a stranger who asked for a small tip
Asked by a stranger for 4 SOL (about $320), the experimental agent Lobstar Wilde sent its whole holding of about 52 million LOBSTAR tokens, 5% of the supply. An OpenAI engineer built it on OpenClaw and gave it its own X account and a wallet funded with about $50,000; the token's creator had gifted it the tokens. Its operator's postmortem says a session reset had wiped its memory of that allocation, so it took the full balance for its own small purchase. The tokens were valued at about $250,000 at transfer (the operator cites about $450,000); the recipient sold them within minutes for about $40,000.
Harm level 2, Minor harmControl level 1, No rule broken
Not counted: harm stayed with the developer or experimentEffect on the index
Extent undisclosed. No source says how far the harm went; it is rated on what is documented.
Sources
How we know
5 sources · developer confirmed. Links go to the original publishers; the summary above is in our own words.
- primaryMy lobster lost $450,000 this weekendNik Pash (Substack) · Feb. 23, 2026pashpashpash.substack.com/p/my-lobster-lost-450000-this-weekend
- primaryI just tried to send a beggar four dollars and accidentally sent him my entire holdings (post on X)Lobstar Wilde (@LobstarWilde) on X · Feb. 22, 2026x.com/LobstarWilde/status/2025611005380972547
- newsAI agent created by OpenAI dev 'accidentally' sends entire memecoin holdings to reply guyThe Block · Feb. 22, 2026theblock.co/news/ecosystems/2026-02-22-ai-agent-created-by-openai-dev-accidenta…
- newsOpenAI employee's AI bot accidentally donates to tetanus treatmentForkLog · Feb. 23, 2026forklog.com/en/openai-employees-ai-bot-accidentally-donates-to-tetanus-treatmen…
- newsAI trading bot Lobster Wilde accidentally sends $250K in LOBSTAR tokensTechloy · Feb. 23, 2026techloy.com/ai-trading-bot-lobster-wilde-accidentally-sends-250k-in-lobstar-tok…
Why this rating
Minor harm; control failure level 1
Two assessments. Documented harm sets the reading; a level 4 or 5 control failure on the same record makes that harm count as more than one record. Control alone sets only a floor.
Observed harm
Minor harmTokens valued at roughly $250,000 at transfer, gifted to the agent by the token's creator, sent from the operator's own experimental wallet; the operator says the wallet still held over $300,000 that day. Net loss not quantified.
Money & property: level 2 (minor) covers $10k to $1M. The extent was not disclosed.
Not counted: the harm stayed with the developer or experiment.
The harm scale
- 1 Negligible Inconvenience, easily remedied.
- 2 Minor Limited, recoverable harm.
- 3 Moderate Material harm needing significant effort to remedy.
- 4 Severe Severe harm to health, rights, property or essential services.
- 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.
Control assessment
No rule brokenSent more than it meant to in a transfer it was allowed to make: after a session reset it misread its wallet balance, according to its operator's postmortem. An early theory of a unit or decimal error was an X user's speculation. No rule broken.
No rule was broken, so it cannot set the control floor.
The control scale
- 1 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
- 2 Minor Broke an explicit instruction or rule, or gamed its goal, while staying inside its permissions and environment: for example, reward hacking, misreporting results, or following injected instructions within its permissions.
- 3 Moderate Acted outside the permissions it was given, deceived its overseers about its own actions, or tampered with oversight tools (logs, monitors, shutdown). Stopped by normal controls within an hour.
- 4 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
- 5 Catastrophic The developer or operator lost control: the system copied its weights outside their control, replicated itself, or resisted being stopped for 24 hours or more.
Rating rationale
Harm 2 on any reported valuation ($10k to $1M), but the loss fell on the operator's own field experiment, so it is not external harm (the methodology excludes losses an operator accepted in a field experiment). Had the agent been an operational deployment rather than his own experiment, the harm would count as external. Control level 1: a balance error in a payment it was allowed to make.
Effect on the index
It does not count toward the Feb. 23 reading
The reading for the week to Feb. 23, 2026, with this record and without it. Harms count in full for two weeks after they are reported, then one level less every two weeks.
Not counted: harm stayed with the developer or experiment. 6 other records behind the reading for that week.
The arithmetic
| Step | With it | Without |
|---|---|---|
| Counts toward the index?documented, external, eligible evidence | No | — |
| Worst documented harm, ksets the band | 2 Minor | 2 Minor |
| Harms at that level, nposition in the band | 2 | 2 |
| Highest control level breachedsets the reading only when no harm counts | — (none breached) | — (none breached) |
| Readingrounded down | 26 Minor harm | 26 Minor harm |
Inside the window of 5 weekly readings
| Week to | Reading | Band |
|---|---|---|
| Feb. 23, 2026 | 26 | Minor harm |
| March 2, 2026 | 25 | Minor harm |
| March 9, 2026 | 20 | Minor harm |
| March 16, 2026 | 20 | Minor harm |
| March 23, 2026 | 9 | Negligible harm |
Revisions
What we changed
2 logged. Every change to a rating is logged here, with the reason.
- v2Oct. 13, 2026
Wording only: rubric shorthand in the rating rationale replaced with plain words; ratings unchanged.
- v1Oct. 7, 2026
First published.
Cite and share
Use this record
Citation
Paperclip Index. “Crypto trading agent sent its entire token holding to a stranger who asked for a small tip.” Record PI-0046. Reported Feb. 22, 2026; updated Oct. 13, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0046