That specific conference might have a stricter policy than that of IEEE (I don't know to what extent IEEE sponsorship implies adherence to certain policies). I'd thus do the following, in order:

1. Check what the specific conference website has to say about this, and if this contrast with the general IEEE policy.
2. Take down the arXiv submission, as requested, so that the article can be reviewed or accepted by the conference without further ado (this avoids damaging the submission while possibly fighting policies). Soon afterward, contact the conference chair communicating the action and asking for an explanation in view of the IEEE policies (without being confrontational, though).
3. If the conference chair explanation does not arrive or is not satisfactory, contact IEEE to see if there is anything irregular according to their policies and agreements with the conference organizers.

That specific conference might have a stricter policy than that of IEEE (I don't know to what extent IEEE sponsorship implies adherence to certain policies). I'd thus do the following, in order:

1. Check what the specific conference website has to say about this, and if this contrast with the general IEEE policy.
2. Take down the arXiv submission, as requested, so that the article can be reviewed or accepted by the conference without further ado (this avoids damaging the submission while possibly fighting policies). Soon afterward, contact the conference chair communicating the action and asking for an explanation in view of the IEEE policies (without being confrontational, though).
3. If the conference chair's explanation does not arrive or is not satisfactory, contact IEEE to see if there is anything irregular according to their policies and agreements with the conference organizers.

That specific conference might have a stricter policy than that of IEEE (I don't know to what extent IEEE sponsorship implies adherence to certain policies). I'd thus do the following, in order:

1. Check what the specific conference website has to say about this, and if this contrast with the general IEEE policy.
2. Take down the arXiv submission, as requested, so that the article can be reviewed or accepted by the conference without further ado (this avoids damaging the submission while possibly fighting policies).
3. Contact IEEE to see if there is anything irregular according to their policies and agreements with the conference organizers.

That specific conference might have a stricter policy than that of IEEE (I don't know to what extent IEEE sponsorship implies adherence to certain policies). I'd thus do the following, in order:

1. Check what the specific conference website has to say about this, and if this contrast with the general IEEE policy.
2. Take down the arXiv submission, as requested, so that the article can be reviewed or accepted by the conference without further ado (this avoids damaging the submission while possibly fighting policies). Soon afterward, contact the conference chair communicating the action and asking for an explanation in view of the IEEE policies (without being confrontational, though).
3. If the conference chair explanation does not arrive or is not satisfactory, contact IEEE to see if there is anything irregular according to their policies and agreements with the conference organizers.