Correct me if I'm wrong.. but, doesn't anything that you actually type into or use the keyboard with need direct access to the.. erm.. keyboard?
Maybe. I wasn't able to understand what specifically trigger that alert but even if you disable direct keyboard access right for ,eg. notepad, it still possible to type text (please test this to sort out the chance there is something wrong with my setup

).
According to
Anti-Keylogger Tester v3.0 direct keyboard access is at least required for GetKeyState,GetAsyncKeyState,GetKeyboardState,GetRawInputData APIs but CFP may trap also other APIs as well.
Those APIs are not malicious by themseves but they could be used for keyloggin purposes.
Opera, Firefox and other applications (even IE on my PC) trigger Direct keyboard alerts, maybe a totally different API is involved.
Such alerts alerts doesn't look relevant enough to guess the app has a keyogging purpose.
IMHO some alerts means that an app has "chances" to be used for keylogging purposes.
If you run Anti-Keylogger Tester v3.0 and deny direct screen access at startup you'll see that the splashscreen will not be displayed correctly.
The corresponding APIs needed for such feature could also be used also to grab a screenshoot (screenshot2 test).
In such cases I guess that the answer to these alerts can be only based on the trust abiut the legitimate purpose of such programs.
Only a RE professional could be able to find out if a program is really malicious.
IMHO it won't hurt to test more restrictive policies to find out if a software really need some access rights.
I really hope that something like a
behavioural fingerprinting standard could be used in future.