On Monday, the Washington Post dropped a bombshell, reporting that Donald Trump had shared highly classified “codeword” intelligence—provided by an ally on the condition that it not be more widely disseminated—with Russian officials during their meeting last week. While administration officials initially issued fierce denials, national security advisor H.R. McMaster, who had himself blasted the story as “false” in a carefully‐​worded statement, effectively confirmed the key elements of the report at a press briefing Tuesday morning. While McMaster repeatedly insisted that Trump’s decision to share information had been “wholly appropriate,” his remarks (perhaps inadvertently) raised several additional grounds for concern.

First, let’s dispense with the obvious: Classification authority in the United States flows from the president, and so a president is legally entitled to declassify or disclose information as he sees fit, for any reason or no reason at all. This is a case where that infamous Nixonism—“When the president does it, that means it is not illegal”—actually applies. Nobody, as far as I can tell, is seriously disputing that. It’s also true that presidents often choose, for strategic or diplomatic reasons, to share particular pieces of intelligence with foreign governments. Yet this does not appear to have been a “routine” instance of such sharing, as McMaster sought to characterize it—not by a longshot.

Rather, as NYU law professor Ryan Goodman observes at the Just Security blog, any decision to share such sensitive information would normally be subject to a rigorous interagency process, allowing the originators of the intelligence to assess the equities implicated by disclosure and apprise the White House of the potential consequences. In this case, McMaster confirmed, the decision appears to have been made on the fly during the course of the discussion—and so necessarily uninformed by any serious analysis of the costs and benefits. Indeed, McMaster even attempted to allay any concerns that Trump might have compromised “sources and methods” by noting that Trump had not been briefed on the source of the intelligence. Yet as intelligence officials so frequently remind us in other contexts, sources or methods can sometimes be reverse‐​engineered from the substance of intelligence. If Trump was not aware of the source, his decision to disclose cannot have factored in either that direct risk of exposure, or the related risk of damaging relations with an ally by sharing sensitive information without seeking permission. Even if he had not been briefed on the details, of course, information shared under such conditions should have been clearly marked “NOFORN” to indicate that it should not be disseminated to foreign nationals, including allies.

Subsequent reporting has identified the source of the intelligence as Israel. This would normally be the most likely candidate, though it sounds slightly odd given the original Post story’s claim that the information had come from a country whose intelligence sharing arrangement with the U.S. was itself a tightly held secret. Possibly this is attributable to an overstatement by the Post, though as a professional paranoid I feel obliged to flag the possibility that it represents an effort at damage control intended to provide cover to another country—Jordan, say—where close cooperation with the United States would be more domestically controversial. (Both could be true, of course, if the information was synthesized from multiple sources—with one source emphasized to draw attention from the other.) Either way, circumventing the usual equities process under the circumstances seems doubly irresponsible, especially given that allies had already expressed skittishness about sharing sensitive information with this administration.

It may well turn out that the substance of what Trump shared was not detailed enough to risk human intelligence sources, at least in isolation, or that it only confirmed information the Russians already had on their own. But in the absence of meaningful internal review prior to disclosure, that would be a matter of luck. What seemed “appropriate” to McMaster might seem like an unacceptable risk to analysts with a clearer sense of what could be derived from that information in combination with other data available to Russian intelligence—either about the extent of our information sharing with allies or about those allies’ own intelligence sources. The point of interagency review is precisely that these things aren’t necessarily obvious on the basis of cursory reflection, even by someone as well informed as McMaster. Thus, even if the information shared should ultimately prove innocuous—though at least some sources are supposedly claiming the reality is actually “worse than has been reported”—the fact that it would be disclosed, essentially, on a whim and without normal process, in violation of an agreement to the contrary, will inevitably (and justifiably) make allies more cautious about future cooperation.

At the same time, this latest in a string of damaging leaks reflecting negatively on Trump, even if it originated from within the White House, seems certain to further strain the already fraught relationship between Trump and his own intelligence community. That’s harmful in the short term, because of course one wants the president to be able to trust the agencies tasked with providing him with information, and the agencies to be able to keep the president fully informed without worrying that sensitive data will be whimsically disseminated. It’s dangerous in the longer run because, as I worried in a previous post, it increases the risk that Trump, feeling besieged, will seek to clean house at the intelligence agencies, replacing the career professionals of the “deep state” with loyalists whose chief qualification is a willingness to serve Trump’s interests, and carry out his dictates, without question.