For regulatory purposes, banks divide their activities in two main categories: banking and trading, whereby banks' assets either belong to the banking book or to the trading book (though sometimes they are switched from one place to the other). Banking book assets and trading book assets can have very different regulatory treatment, even if the nature of the asset turns out to be exactly the same.
The trading book was devised to house market-related assets (derivatives, bonds and so on) rather than traditional banking activities. Trading book assets are supposed to be highly liquid and easy to trade. As banks significantly began to grow their market activities since the early 1990s, the trading book became bigger. Regulators decided to let banks use Value at Risk models to calculate capital charges for the trading book. Given how low VaR can be, and given how much banks can love leverage, this gave institutions an incentive to park as many assets as possible in the trading - rather than banking - book.
In the run-up to the 2008 credit crisis, many banks had placed a lot of assets that looked of a banking book-type into the trading book, possibly to take advantage of lower capital requirements. So the big story of the crisis became how banks had accumulated all those billions of subprime CDO tranches (whose eventual meltdown killed the banks) in their trading books. After the crash, regulators recognised that they had been very lax when it came to policing the trading book, permitting banks to abuse it with undesirable results. A thorough review of the trading book is currently taking place, with regulators eagerly trying to avoid future repeats of such pernicious regulatory arbitrage.