The stress acting on fractures in a 3.5 km deep borehole in granite is examined to place constraints on fracture strength and to evaluate whether permeable fractures tended to be critically stressed. Two data sets consisting of some 500 fractures are analyzed. The first considers the permeable/impermeable fracture populations in their natural state. The second considers the populations after the hole had been subject to a prolonged, 9 MPa overpressure in two major injections which increased the number of permeable fractures from 18 to more than 95. For both data sets it is found that permeable fractures are critically stressed inasmuch as they support levels of shear stress that would be verging on failure if their strength were governed by a Coulomb friction law with a coefficient of 0.6--1.0. However, the vast majority of impermeable fractures are also critically stressed, implying that critical stressing is a necessary, but not a sufficient, condition for permeability to develop. Most permeable fractures showed evidence of shear failure after the injections and tended to occur in hydrothermally altered cataclastic shear zones, suggesting that they may have been relatively weak because of the presence of illite. To prevent failure of the impermeable fractures during the injections requires a cohesion of at least 6.5 MPa to augment a maximum allowable friction coefficient of 1.0. These are probably early mode 1 fractures that are sealed primarily with calcite. The strength of the shear zones appears to be significantly greater than the illite-rich fractures from which they are composed. This imposes constraints on their internal architecture.