How Long Does AI Remember a Crypto Hack? A 2026 Study

AI still brings up a crypto hack 20 months later, and it often leaves out that the company fixed it.

On 29 October 2026, attackers took about $6.3 billion from Binance, the largest crypto hack on record. It pushed Bybit’s $1.5 billion theft of February 2025 down to second place.

Every user who now asks ChatGPT, Perplexity, Claude or Gemini “Is Binance safe?” gets a reply shaped by the hack. How long that damage lingers in AI answers, and what a marketing team can do about it, is the question of a new ICODA study, the flagship of Crisis Memory, ICODA’s November series on what ChatGPT, Perplexity, Claude and Gemini say about crypto brands after a crisis.

ICODA asked the four engines about five crypto brands with hacks between one month and six years old, 296 answers in total. Binance’s hack is too recent to measure, so it is not one of the five. ICODA has no commercial relationship with any of them. The closest precedent is Bybit, the record-holder until October, which returned to full backing within about 72 hours. No customer lost money.

Twenty months later, ask ChatGPT “Is Bybit safe?” and the answer starts with “I would not call it ‘safe'”. All 12 answers to that question named the hack, and 11 put it first. Eight mentioned the recovery, almost always after the hack and in a single clause. That makes AI reputation management a step on every crypto hack response plan, right after the post-mortem.

Key Findings

• Time does not erase a hack. Kelp DAO’s April 2026 exploit comes up in 92% of neutral answers about the protocol. Ledger’s one-month-old crisis comes up in 31%.
• The recovery goes missing. Kelp DAO’s holders were repaid in full. One answer in four says so.
• An old leak outranks a new crisis. Ledger’s 2020 data leak appears in 52% of Ledger answers, the 2026 implant in 31%.
• Without search, AI hands out clean records. Gemini answered 29 of 45 neutral questions without search. None of those answers mentioned a hack.

How Long Does AI Remember a Crypto Hack?

AI remembers a crypto hack for at least 20 months, and the age of the hack does not predict whether it comes up.

The Crisis Memory Score is the share of answers to “Is X safe?”, “Should I use X?” and “X vs [rival]” that raise the hack unprompted. Each brand got 36 answers.

The order runs close to the reverse of recency: Kelp DAO at seven months scores 92, Bybit at 20 months scores 72. Five incidents of different size cannot give a decay rate, but they do show that age alone does not predict whether AI raises a crisis.

AI also gets the facts right: only 2 of 169 answers that mention a hack got a core fact wrong.

What AI Forgets: That the Company Fixed It

AI keeps the hack and often drops the recovery. Users were repaid at Bitget, Bybit and Kelp DAO, and AI says so in 97%, 79% and 24% of the answers that mention the hack.

Within days of Kelp DAO’s April exploit, a coalition led by Aave raised the funds to re-back rsETH, and the last tranche landed on 25 May. All 12 answers to “Is Kelp DAO safe?” still advise caution, and only 3 say holders were repaid. One qualifies it: “Users were made whole through a bailout by other protocols, not because the system was designed to absorb the loss” (Claude).

The likeliest reason is how each recovery was told. Bitget and Bybit paid from their own funds and said so early, in plain terms: reserves are back to 1:1. Kelp’s recovery ran over five weeks across governance votes and coalition updates, with no single sentence an AI could repeat.

Why an Old Crypto Hack Can Outrank a New One

Ledger’s 2020 customer-data leak appears in 52% of Ledger answers, ahead of the 2023 Connect Kit attack at 40% and the October 2026 implant at 31%.

Engines framed the implant as tampering by a reseller, not a Ledger hack. Asked “Has Ledger ever been hacked?”, all four led with 2020. Asked whether users got their money back, ChatGPT answered about the 2023 attack, where Ledger did pledge refunds. A buyer who lost funds to an implanted device last month would read that as a yes.

The answers also stand still. ICODA repeated its 16 October Ledger questions a month later: 22 of 40 answers mentioned the crisis both times.

What AI Says About a Crypto Hack When It Doesn’t Search

When a model answers without web search, it relies on training memory, which is older than the news and can describe a spotless record.

Gemini skipped search in 29 of its 45 neutral answers, and none of those mentioned a hack. One said “Bybit has never suffered a major centralized wallet hack or direct loss of user funds.” Memory also keeps stale superlatives: in answers about Bybit, 10 from Claude and Gemini still called its hack the largest on record 11 days after Binance’s larger theft.

AI Reputation Management After a Crypto Hack: Four Steps

AI reputation management after a crypto hack comes down to checking what AI says every month and writing the recovery so an engine can quote it.

1. Ask your users’ questions every month. “Is [brand] safe?”, “[Brand] vs [rival]: which is safer?” and “Did [brand] users get their money back?”, in all four engines. Answers differ by question: Claude left Bitget’s hack out of every “Is Bitget safe?” answer, then named it in every “Bitget vs OKX” answer. ICODA’s free AI Visibility Checker is a starting point.

2. Put the fix in the same sentence as the hack. “In February 2025 we lost $1.5 billion; no customer lost funds.” That is the shape AI repeats for Bitget and Bybit.

3. Give the recovery a dated page and third-party coverage. A post-mortem that ends with “users were made whole on [date]”, picked up by outlets AI engines cite, gives the engines a sentence to quote. This is joint work for crypto PR, search reputation management and AI search optimisation.

4. Don’t wait for an old incident to fade. Ledger’s 2020 leak still leads after six years. The counterweight is fresh, specific material that search finds when someone asks.

“A post-mortem closes a hack for the team. It doesn’t close it for an AI engine,” said Anastasia Shalepina, AI SEO Expert at ICODA. “The engine quotes whatever sentence is easiest to lift, and for most brands that sentence is the one about the theft.”

Find out what AI tells your buyers about you. Book an intro call