Looks like the data line is shorted to a logic signal that's high, so the high level is OK but the low level is only halfway to where it should be.
As an experiment, try putting an additional 3k3 resistor between SDA and GND. This should, in theory, create a potential divider that puts it at ~2.5V. My guess is it'll remain high.