Electron paramagnetic resonance (EPR) signals generated by ionizing radiation in touch-screen glasses have been reported as useful for personal dosimetry in people accidently exposed to ionizing radiation. This article describes the effect of light exposure on EPR spectra of various glasses obtained from mobile phones. This effect can lead to significant inaccuracy of the radiation doses reconstructed by EPR. The EPR signals from samples unexposed and exposed to X-rays and/or to natural and artificial light were numerically separated into three model spectra: those due to background (BG), radiation-induced signal (RIS), and light-induced signal (LIS). Although prolonged exposures of mobile phones to UV light are rather implausible, the article indicates errors underestimating the actual radiation doses in dose reconstruction in glasses exposed to UV light even for low fluences equivalent to several minutes of sunshine, if one neglects the effects of light in applied dosimetric procedures. About 5 min of exposure to sunlight or to light from common UV lamps reduced the intensity of the dosimetric spectral components by 20-60% as compared to non-illuminated samples. This effect strongly limits the achievable accuracy of retrospective dosimetry using EPR in glasses from mobile phones, unless their exposure to light containing a UV component can be excluded or the light-induced reduction in intensity of the RIS can be quantitatively estimated. A method for determination of a correction factor accounting for the perturbing light effects is proposed on basis of the determined relation between the dosimetric signal and intensity of the light-induced signal.