Modern cables don't have high voltage to worry about. The signals are amplified optically:

https://en.wikipedia.org/wiki/Optical_amplifier#Doped_fiber_...

It isn't necessary to bend a fiber either. Optical fibers can be tapped via evanescent wave coupling. The fiber cores just need to be near enough to each other over a long enough span, possibly with some removal of the outer cladding.

Notable patents:

"There is a need for coupling light into and from an optical transmission line without interrupting the flow of transmission. Such lines may be used for communications, as well as other purposes.[1]"

1. https://www.google.com/patents/US3933410

2. https://www.google.com/patents/US4097117

3. https://www.google.com/patents/US4377403

4. https://www.google.com/patents/US4763272

5. https://www.google.com/patents/US4336047

Raman amplifiers still require juice to run the pumping laser. There aren't exactly 110V sockets sitting on the ocean floor.

There is a copper tube around the fiber bundle that carries high voltage (usually 10,000+ volts to reduce losses over long spans). The return is usually just coupled directly to the ocean.

If you're careful, use a non-conductive cutting blade, and have enormous balls, you can no doubt cut through the tube on one side and not cause any noticeable effects. At least, if you've brought along someplace dry to work on it...

Yeah it's just not that easy with the advent of coherent optical communications in subsea networks. You need to be able to do some very complex DSP in real-time at 50Gb/s. You also need to know the exact FEC encoding scheme, grey map, actual modulation format etc... It is very non-trivial. For OOK signals I think it's somewhat doable for coherent, very, very hard. Not totally impossible but I highly doubt even CIA has this expertise.

"Where there's a will (and enough black project money), there's a way."

With love, NSA