Last week, a group of 17 countries along with leading social media platforms signed Prime Minister Jacinda Ardern’s Christchurch Call “to eliminate terrorist and violent extremist content online”.

The non-binding three-page document is inspiringly pragmatic, combining aspirational actions without falling for heavy-handed (and misguided) regulations.

Kudos to the international initiative. There is indeed much to celebrate about such a concerted effort against the spread of online radicalism.

But we must not lose sight of the implementation hurdles ahead – and keep working for a safe and free online community.

Social media uploads of harmful content are not easy to detect. (Nor are there easy answers to philosophical questions about what constitutes objectionable content – and who decides it.)

The sheer number of social media users – Facebook alone has more than 2 billion active users – indicates the solution ultimately rests on more accurate computer algorithms.

Unfortunately, despite already impressive progress, state-of-the-art machine learning technology has a long way to go. Even in the face of flagrant violations of codes of conduct, such as livestreaming terrorism acts, our machines are still learning.

Until artificial intelligence catches up, a viable response is for online providers and governments to learn from past mistakes – which the Christchurch Call rightly alludes to. But this can only go so far, as ill-intended users will always have the first-move advantage.

Facebook, for instance, just introduced a one-strike policy, which temporarily bans users starting on their first offence. Had this policy been in place on 15 March, it would have prevented the terrorist from using his live account that day in Christchurch.

But such a rule still cannot prevent would-be terrorists from gaming the system in the future. Cognisant of Facebook’s new policy, rogue actors could simply fly under the radar on the social media in the weeks leading to their livestreaming of atrocious acts.

And even if technology improves and new codes of conduct minimise risks, there are still plenty of non-compliant online platforms to disseminate terrorism content.

All these hurdles might appear like we are fighting a losing whack-a-mole game. Such is the nature of the fight against online terrorism.

Still, we should not let those obstacles paralyse coordinated efforts, and we most certainly should not fall for easy-but-wrong populist responses.

In that sense, the Christchurch Call is already a clear win.