Fifty one percent attack

An entity controlling a majority of hashrate can, while it sustains that majority: reliably out-build the honest chain to reverse its own recent transactions (double-spend), orphan other miners' blocks, and censor transactions by never including them and rejecting blocks that do. It cannot steal arbitrary coins, mint extra coins, or alter consensus rules — every non-mining node still enforces validity. The attack is also expensive to sustain and self-limiting: it degrades confidence in exactly the asset the attacker's hardware and block rewards are denominated in.

Nuance: "51%" is a threshold of reliability, not possibility — reorg attacks succeed probabilistically below 50% too (whitepaper Section 11 quantifies this), which is why high-value settlement waits more confirmations. Smaller proof-of-work chains have suffered real 51% double-spend attacks; Bitcoin's defense is the sheer cost of matching its hashrate, an economic fact that drifts and should be quoted with a date if quantified.

Common misstatements:

• "A 51% attacker controls Bitcoin." — They control block production while the attack lasts; validity rules, existing balances, and issuance remain intact.
• "51% attacks are impossible on Bitcoin." — They are economically prohibitive, not physically impossible; the claim should be stated as cost, not impossibility.
• "A government majority attack would end Bitcoin permanently." — Speculative; responses (waiting more confirmations, ignoring attacker chains socially, difficulty/PoW changes as last resort) are debated countermeasures, not settled outcomes.