David,
It looks like this is a free beam laser (not fiber optic coupled). If you are planning to go into fiber optics as implied by your first line, I would start out with a laser diode that is pre-coupled. Almost all of your loss is in the coupling, not the fiber optic cable itself. In glass fiber optics, loss is often quoted per kilometer. The alignment of a diode and fiber is challenging and ranges from 50-85% efficient when done commercially. You will need to work hard to match this efficiency without the advantages of a specialized shop.
Eye safety is discussed extensively in the ANSI Z136 standard (you probably want Z136.1). The key is power density (mW/cm^2 or mJ/cm^2). This beam is collimated (more dangerous as it does not spread out) and ‘continuous wave’ (not pulsed, safer as you don’t need to worry about peak power, only average). Echoing what John Marlow stated above, laser class is determined by the entire packaged product, NOT what you can get at with a screwdriver, etc. If it’s a Class 4 inside a locked box and attenuated so that the output is Class 1, it can be sold as a Class 1. You are opening the box and caution is justified.
Note that laser light coming out of a fiber optic is not collimated. The numerical aperture of the fiber (NA, in air, sine of half angle of light emission) determines how the light spreads out. The further away from the fiber tip you are, the lower the power density (more area, same amount of optical power). In the Class 3b devices I use with 0.22 NA fibers, the power density drops down to eye safe levels in a short distance (for my wavelength/power/NA combination, 10-15 cm. NOT a rule of thumb).
Side note: although this is a 532nm laser, there may be other wavelengths present as alluded to above. There was an article a few years ago (Physics Today?) describing an experiment for high school kids with 532nm laser pointers and a CD as a grating. These lasers are often frequency doubled 1064 nm (not visible). The 1064 pump is much more powerful than the visible output. In inexpensive lasers, the pump beam was found to be frequently not filtered out of the output. As long as these are co-linear, the eye risk is partially mitigated by the blink response of the eye (to the green light). As soon as the light goes through a lens/prism/etc, the 1064 and 532 nm are no longer aligned. I unfortunately don’t have the reference for this handy.
If you want to discuss the overall design, I might be able to give you some suggestions on the light source to reduce the potential safety concerns. For example, in low power applications, even though the coupling efficiency into fiber optics is poor, LEDs may make more sense.
r/
David