There has been a push for contact-rich manipulation in robotics, where meaningful and deliberate contact with the environment is required. By “meaningful,” we refer to purposeful interactions such as pushing a button, inserting a humidity sensor into soil, probing the temperature of a cake, or squeezing a bottle to grasp it. These tasks highlight the need for contact while demanding strict safety guarantees to prevent damage to objects, humans, or the robot. Traditional robot safety assumes contact should be avoided altogether, which constrains contact-rich tasks. As a result, controllers often operate with either reduced safety margins or with reduced velocity, thus decreasing the capabilities of the robot. An ideal controller operates the robot within the full safety margins and with minimal restrictions to its capabilities. This paper uses Control Barrier Functions (CBFs) to ensure that contact with a surface is below a maximum safety velocity and the subsequent surface penetration is bounded, while allowing free movement when contact is not expected. We derive an exponentially decaying velocity–distance CBF and test it in a torque-controlled manipulator simulation. The simulation shows the manipulator slowing down to a safe contact speed and not exceeding the maximum allowed penetration, even when the nominal controller is unaware of the safety constraints. The results are promising and open the door to additional research in safety guarantees in contact with Control Barrier Functions.