There may be bug (or luck to find some number) and then creating bitcoins out of nothing. And nobody can verify.
That is the risk - the scheme relies on the security of ECDSA to protect the currency supply.
We're already relying on ECDSA to protect our balances from overt theft, so I'm not sure how much it actually changes the security model to also rely on it to protect our balances from covert theft via counterfeiting.
The reason I'm interested in amount blinding is that my current project is working out a multi-step plan to kill graph analysis. With the right plan and without blinded amounts we can kill graph analysis, but with blinded amounts we can drive a stake through its heart to make sure it stays dead.
We are also using SHA-256 and RIPEMD-160 hashes to protect our balances. So even ECDSA is broken our balances can be safe and then ECDSA replaced.