On-line process control equipment for CO2 laser cutting is not available for industrial applications today. The majority of the industrial cutting machines are regulated off-line by highly-educated staffs. The quality inspection of the samples often is visual, and referred to different quality scales. Due to this lack of automatization, potential laser users hesitate to implement the cutting method and hereby to benefit from the advantages offered by the method. The first step toward an automatization of the process is development of a process monitoring system, and the investigation described in this paper is concentrated in the area of on-line quality detection during CO2 laser cutting. The method is based on detection of the emitted light from the cut front by photo diodes. The detection is made co-axial with the laser beam to assure independence of the chosen processing direction. ZnSe mirrors have been placed in the beam path, reflecting the laser beam but transmitting the visible light emitted from the process. Cut series of 2, 6 and 8 mm mild steel have been performed. Fourier Analyses and statistical analyses of the signals have been undertaken, and from these analyses it is possible to estimate the surface roughness in the cut kerf, dross attachment at the backside of the work piece and the penetration of the laser beam.