There is a classical correspondence between Lie algebras (over $\mathbb{R}$ or $\mathbb{C}$) and Lie groups in the finite dimensional case: to every Lie group $G$ there is an associated Lie algebra $\mathfrak{g}$, and conversely. Moreover, this correspondence is one-to-one if one requires $G$ to be simply connected. One also has maps $\exp$ and $\log$ that map between a Lie group and its associated Lie algebra, and the Baker–Campbell–Hausdorff (BCH) formula gives a Lie series $z$ for given $x$ and $y$ in $\mathfrak{g}$ such that $\exp x\exp y=\exp z$ (at least for sufficiently small $x$ and $y$).
Every Lie group gives rise to a Lie algebra, but in the infinite-dimensional case there are ‘non-enlargeable’ Lie algebras which don't correspond to a Lie group.
However, I came across a remark in Does a finite-dimensional Lie algebra always exponentiate into a universal covering group saying that every Lie algebra over $\mathbb{R}$ can be exponentiated to give an abstract group, but that it is a non-trivial theorem. Rather than poking this old comment to another question (where I may have taken something out of context), I thought this deserves to be explicitly asked.
Let $\mathfrak{g}$ be a Lie algebra over a field $k$. Is there an abstract group $G$ with which $\mathfrak{g}$ is naturally associated? What if we restrict to $k$ having characteristic zero — or even $k=\mathbb{R}$ or $k=\mathbb{C}$?
By ‘naturally associated’ I am thinking of a bijective correspondence reminiscent of the classical situation, where the algebra and the group are linked by the BCH formula, and Lie algebra homomorphisms $\phi$ lift to group homomorphisms $\Phi=\exp \phi\log$. However I am not expecting a differentiable structure on $G$. (That said, if $G$ comes with some extra structure for free, I would be interested to hear about it.)
I have looked at other MO questions such as Lie Groups and Lie Algebras but it's not clear to me that infinite dimensional Lie algebras are considered there.
A further observation is that the Lie series encountered in the classical case converge in the usual sense on sufficiently small neighbourhoods. However, in some cases of interest to me, such as free Lie algebras, the Lie algebra can be embedded in an (associative) algebra of formal power series with non-commuting indeterminates, in which case there are no restrictions on when the Lie series converge. So it seems to me that a subcase of the question above corresponding to the situation where all Lie series converge may well have been studied. There is a 1948 Doklady paper On normed Lie algebras over the field of rational numbers by Mal'cev which sounds relevant but has proven resistant to my attempts to track it down.
I would be grateful for any pointers to books or articles that discuss this.