The BBC discusses a man-in-the-middle attack on EMV payment cards, also known as Chip-and-PIN. The attack was developed by a team at Cambridge University in the UK. Using a real card wired up to a laptop, connected to a fake card that is inserted in the POS terminal, this attack can authorize payments with an arbitrary PIN.
I am not sufficiently familiar with the cryptography in use for the EMV protocol, but my first thought is that astute observation by POS personnel should provide substantial defense against this entire class of attack: if someone shows up at your cash register with an EMV card wired to his backpack, something fishy is probably afoot.
My other thought has to do with the notion that banks might attempt to shift the responsibility for fraudulent Chip-and-PIN transaction to the consumer. “Since EMV is so secure,” the reasoning goes, “the PIN authorization is proof positive that the transaction is valid.” Except it has now been shown that PIN authorization can be spoofed.
Security is not black and white: it does not make fraud impossible, but makes it harder and more expensive to commit fraud. The protection level provided by a security feature should be commensurate to the value of the transaction it protects. Too high a protection level is likely to be more cumbersome, or more expensive, than the transaction in question justifies.