Sam Sloan wrote:
You can check all the databases. There is no game with the opening
moves 1. d5 Nf6 2. c4 c5 3. d5 b5 in any database anywhere, prior to
when Benko started playing it.

Actually, I watched the first game with the Benko Gambit as it was
played. The game was Laver-Benko, American Open, Santa Monica 1967.

Sam is correct that no database has any game beginning 1.d5, at least
none of good repute.
However, assuming he meant 1.d4 Nf6 2.c4 c5 3.d5 b5, there are many
examples before 1967. The earliest I found in Chess Assistant 5.0 is
Morales-Vaitonis, Buenos Aires 1939; next is Alatortsev-Ratner, Moscow
1945.
If we refine the definition of the Benko Gambit to the point where a
gambit is actually offered, after 4.cxb5 a6, CA's first example is
Szabo-Lundin, Saltsjobaden Interzonal 1948, still 19 years before 1967.
Therefore Sloan is completely wrong to claim that there is no such
game "in any database anywhere." I doubt he even bothered to check very
many (or any?) databases before making his absolute claim, but that is
typical of Sloan's research methods.

The Oxford Companion points out that the line was played in the 1930s
by Opocensky, and "used occasionally from 1947 by Bronstein [and]
Keres." However, the OC also supports Benko's name being associated
with line: "The gambit's acceptance as a standard line owes much to
Benko who published analysis and wrote a book, 'The Benko Gambit'
(1973)."