A recent Financial Times profile of the U.K.’s newest scientific innovation office has garnered attention, and rightly so. Modeled after the U.S.’s Defense Advanced Research Projects Agency (DARPA), ambitions for the U.K.’s ARIA (Advanced Research and Invention Agency) are high. Its aims are nothing short of transforming science by stimulating revolutionary technologies. Yet the foundation the agency is built upon is already beginning to show some cracks.

Ilar Gur is the new chief executive of ARIA, and he has clearly indicated that the agency’s destiny will be significantly determined by its eight newly-appointed program directors. With each of these directors wielding control over a formidable £50 million purse, a lot hinges on these individuals having not just good, but exceptional, judgment. One would hope therefore that an impeccable hiring process was followed. However, the first ARIA chief executive stepped down before even beginning the role, hinting at administrative challenges.

Gur’s stance is clear: ARIA’s directors are encouraged to make bold and risky investments. But this audaciousness—while commendable on some level—is fraught with peril, especially given the absence of clear metrics to gauge success. ARIA has a long-term vision, spanning decades, but it will need a precise roadmap of where it is going and how it will get there, or else the nascent agency is likely to lose focus.

Envisioning a government entity with the risk appetite and adaptability of a venture capitalist sounds innovative on paper. Yet history paints a very different picture. Governments’ attempts to stimulate specific sectors through “industrial policy” often go awry. The U.K. experience with the Concorde supersonic aircraft and the nationalized automotive company British Leyland stand as glaring reminders of ambitious projects that ended up draining taxpayer money.

Another concern is the potential for industry capture. The financial and pharmaceutical sectors’ evident interest in ARIA raises alarm bells about the possibility of conflicts of interest, especially given the likelihood of revolving doors between government and corporate jobs.

While DARPA in the United States has had its share of wins, it’s important to remember its numerous missteps as well. The Strategic Computing Initiative from the 1980s, which aimed at AI advancements, was largely a flop. Other U.S. agencies modeled after DARPA, like ARPA-E and ARPA-H, have either fallen short of expectations or it’s too early to tell if they are succeeding.

DARPA’s substantial achievements, in reality, are probably a byproduct of the enormous amounts the U.S. Congress spends on defense each year. Historically, governments have always been fairly innovative when it comes to warfare, making DARPA’s “success” less exceptional and more expected. With such extensive financial backing, a lack of any concrete results would have been a bigger anomaly.

Though DARPA did fund early stages of the internet, the internet’s development was probably inevitable. And although government contracts were important for some of Silicon Valley’s earliest companies, this is not really evidence of the public sector’s venture capitalist abilities. Government at all levels accounts for forty cents out of every dollar spent in the U.S. economy, so nearly any successful venture is going to have public sector connections somewhere.

Sadly, there are many instances where government funding of research often diverges into the absurd. Who can forget the study of cocaine-snorting monkeys funded by the 2010 stimulus law? (Subsequently, they moved on to nicotine.) Each year small-government legislators produce reports chock full of examples of government waste, and scientific research funding routinely tops the list. Is yet another government research agency really the best way to stimulate innovation?

Unlike private sector ventures motivated by profit, government agencies inherently lack urgency. This helps explain why some are calling ARIA’s first few months “lethargic.” A potential solution lies in aligning director salaries with project successes, and adopting an approach to investments whereby government takes an equity stake in the risky companies it finances. Beyond aligning incentives, that way taxpayers won’t just be left holding the bag when risky ventures don’t pan out—they can capture a share of the proceeds too, when projects succeed.

ARIA’s current model places excessive reliance on directors. Even if it succeeds in finding the right leadership today, there’s no assurance of sustained excellence. Solid, measurable metrics should be the driving force, superseding mere hope of finding the right people.

ARIA may also be too focused on short-run fads and making headlines. Apparently, the agency plans on investing significantly in artificial intelligence. That may help Gur and his colleagues land profiles in the Financial Times, but AI is already receiving plenty of private sector backing given the massive media attention the technology is receiving. Contrast ARIA’s model with sovereign wealth funds—another government investment vehicle—which tend to operate more quietly and conservatively, looking for hidden value in unexpected places and buying at unpopular times, all while seeking to avoid the media spotlight.

While the U.K. government’s foray into venture capitalism is ambitious, it’s a path riddled with potential pitfalls. British taxpayers deserve a robust, accountable, and clear strategy. As it stands now, ARIA risks starting off on the wrong foot. A course correction should be non-negotiable.