Check your connections on your cables. A bad connection can cause this, also a bad cable is possible.
If you are using a clamp on to measure this that can be fooled by being too close to an adjoining phase wire.
Check the stupid things first, then proceed to the less likely causes.

Thanks Cranky108. The current reading is OK as the cable drawing the higher current is hotter.

I'd also suspect a bad, high-resistance termination as already mentioned. Cable lengths should be approximately equal. The shorter the cables the greater the impact of cable length on division of current. A low-resistance ohmmeter can be used to check cable circuits to find differences in end-to-end resistance.

All the cables enter through a steel gland plate. We checked for loose connections but could not find one. Could Eddy current be the cause of the current unbalance? The steel gland plate is quite hot.
Guardiano

How many entry holes in the plate?

Dear Mr. Guardiano (Electrical)(OP)18 Feb 22 14:43
1. "... We are commissioning a 2000 k VA generator and the current on each phase is balanced. There are four single cores cables per phase but the current on each core is not balanced at all. The cables are run in trefoil. What could be the cause?..."

Check that the "trefoil" is arranged [viewed ccw phase sequence] (ABC), (BCA), (BCA) and CAB).

2. "...Could Eddy current be the cause of the current unbalance? The steel gland plate is quite hot..."
When the conductor current say >400A, eddy current heating comes into the picture.
2.1 Proposal:
a) with existing metallic gland plate: cut a thin slot say 0.5 mm between all holes into two semi-circles. Note: the gland plate will be in three pieces with 0.5 mm air gap,
b) replace the existing steel gland plate by [brass, aluminum or stainless steel] material.
Che Kuan Yau (Singapore)