By that argument, any chain with less hashing power than Ethereum (subtracting out ASIC miners that can’t mine a different chain) is insecure. Similarly, we could assert that any chain that could be attacked by someone spinning up enough AWS/Azure/GCE nodes (i.e., there exists enough access to global hashing power) to attack it is similarly insecure.
If that is your definition of security, then I will agree to your assertion that a chain split results in one of the two chains being insecure. However, I strongly argue that this isn’t how chain security is (should) be measured. What matters is whether or not there exists enough financial incentive to attack the chain such that one can profitably do so.
In the example of diverting hashing power to attack some chain, unless you have an attack that will allow you to double spend more than it costs you to divert your hashing power then I argue that the chain is secure. The recent double spend attacks against blockchains have occurred because exchange platforms have allowed users to transact for larger values than the system can properly secure against. This is a mistake by exchanges, they should never be transacting off-chain for more than the security of the system (which to be fair, is hard to measure).