We present a new determination of the large-scale clustering of the CIV forest (i.e. the absorption due to all CIV absorbers) using its cross-correlation with quasars in the Sloan Digital Sky Survey Data Release 12. We fit a linear bias model to the measured cross-correlation. We find that the transmission bias of the CIV forest, , at a mean redshift of |$z\, =\, 2.3$|⁠, obeys |$(1\, +\, \beta _c)b_{F c}=-0.024\, \pm \, 0.003$|⁠. Here, β_c is the linear redshift space distortion parameter of the CIV absorption, which can only be poorly determined at β_c = 1.1 ± 0.6 from our data. The most accurately determined combination marginalized over β_c is |$(1\, +\, 0.44\, \beta _c)b_{F c}=-0.0170\, \pm \, 0.0014$|⁠. The transmission bias is related to the bias of CIV absorbers and their host haloes, bτc, through the effective mean optical depth of the CIV forest, which we estimate at |$\bar{\tau }_c(z) \simeq 0.01$| from previous studies of the CIV equivalent width distribution. We then find 1 < bτc < 1.7, with the large error arising from uncertainties in βc and |$\bar{\tau }_c$|⁠. This CIV bias is lower than the DLA bias bDLA ≃ 2 measured previously from the cross-correlation of DLAs and the Ly α forest, indicating that most CIV absorbers are hosted by haloes of lower mass than DLAs. More accurate determinations of |$\bar{\tau }_c(z)$| and βc are necessary to check this conclusion.