As other commenters have mentioned, MIT paying for articles a la carte worked out cheaper than getting all-access subscription from Elsevier. Presumably Elsevier would want to stem this sales channel so institutions would be forced to pay more.
Any individual publisher which restricted access via these brokers would face a sudden loss of revenue and a knock on effect to the prestige of their journals. If they allowed some customers to use these services but not others, the predatory nature of the model would stick out a mile to institutes and regulators.