It basically is already. In the US, if you produce a peer-reviewed paper resulting from federal funding you must provide "Public Access". Each agency has their own way of implementing it. For example, a publication funded by NIH must be uploaded to pubmed.
Search for Public access policy and agency name.
Perhaps it says that on paper, but in practice it is not enforced in many areas. Published papers funded with government grants are behind paywalls; data and computer code are routinely not archived in a publicly accessible location. The example you give, PubMed, is actually the exception, not the rule, in practice.
Perhaps I should have said "required by law, which is actually enforced".
How much do publishers really need to do their job well? Just pass that law and you'll quickly find out by way of the fees charged to publish.
And it seems odd for the solution to be “force Elsivier to make articles free”. That’s like saying “Harvard and Stanford have the most prestigious degrees, but they charge too much, so the only fair thing to do is to force them to make it free”.
Seems to me the main problem is universities place too much emphasis on these “high impact” journals when it comes to career progression. Measure performance a different way and the problem is solved.
But, everybody doesn’t want to let go of what they get out of it, they want everyone else to give up their benefits.
Researchers - love the prestige of getting published in Science, hate that their articles aren’t freely available to other scientists (want their cake and eat it too). Could publish in open access journals but don’t want to take a hit in “impact factor” (first mover disadvantage).
Universities - love that someone else gets to judge the performance of their researchers rather than them. Hate that they spend tens of millions on access, but hard to say “publish in Science but we won’t pay for access” (cake and eat it too)
Journals - love the fact they can get fantastic papers submitted, and free peer review solely because of the prestige built up over decades/centuries. Want to charge the “willingness to pay” price for the prestige they bring, but cant justify the price versus cost of publication.
If you think the cost of a subscription is high for universities, which have huge endowments that are invested in all kinds of assets (I have seen it commented that major universities in the US are really hedge funds that happen to do teaching and research on the side), imagine what it's like for individuals with no academic affiliation who just want to see what their tax money is actually funding.
In other words, the "sweetheart deal" is being able to afford the subscription at all, given that in practice open access does not happen.
Then why call it a “deal”?
Because compared to what an ordinary taxpayer gets, it is. And that's the relevant comparison since ordinary taxpayers are the ones ultimately paying for the research.
Not if they want to have their academic jobs and get grant funding for their research. The entities that have sweetheart deals with companies like Elsevier aren't scientists; they're universities and governments. And universities and governments control academic jobs and grant funding.
The real fix is for it to be required by law that any research that was funded by government grants must be published with open access.