Allowlist

An allowlist is an enumeration of things that are explicitly permitted. Anything not on the list is thereby automatically forbidden. Its counterpart is the blocklist: there, what is forbidden is listed, and everything else is permitted. The difference sounds small, but it is decisive. With an allowlist, one must know in advance what one wants to permit. With a blocklist, one must add every new danger individually as it appears. In German, the allowlist is called “Positivliste,” and the blocklist “Negativliste.”

Why allowing is safer than forbidding

The advantage of an allowlist is its behavior toward the unknown. If something entirely new appears, it gets blocked, because it is not on the list. A blocklist, by contrast, lets it through as long as no one has entered it yet. Experts call this principle “deny by default”: when in doubt, permit nothing.

This is precisely where many protective mechanisms fail. A classic antivirus program works with a blocklist of known malware. Anyone who distributes brand-new malicious software circumvents this list until the manufacturer becomes aware of it. A company computer that is only allowed to launch twenty approved programs is immune to that same malware. This is why security agencies mandate allowlists for critical systems such as power plants or hospitals.

The price for this is convenience. Every allowlist regularly blocks harmless things too, simply because they are missing. This is called a false alarm. Anyone operating such a list therefore needs a process for how new items are reviewed and added. Without this process, the list gets circumvented or shut down entirely.

What is on the list and who maintains it

An allowlist needs a clear characteristic by which it recognizes entries. For programs, this is often a digital fingerprint of the file, a so-called hash. If the file changes by even a single character, the fingerprint no longer matches. For websites, it is addresses or entire domains; for emails, sender addresses; for networks, IP addresses.

The real work is not the technology but the maintenance. Someone must decide who is allowed to submit a request for inclusion and who approves it. Large organizations combine this with software distribution: approved programs are installed centrally and thereby automatically end up on the list. If a list grows without scrutiny, it loses its purpose. An allowlist with ten thousand entries ends up permitting almost everything.

A common misconception is to view allowlists and blocklists as opponents. In practice, both are combined. A mail server can reject known spam senders via a blocklist while simultaneously always letting important business partners through via an allowlist. The allowlist usually takes precedence in such cases.

Allowlists in AI systems and the news

In AI applications, one encounters the term repeatedly. A language model permitted to research on the internet is often given an allowlist of trustworthy sources. It then reads only specialist databases or vetted news sites. This lowers the risk of it adopting misinformation or reacting to hidden instructions on a manipulated website.

Allowlists are also standard for AI agents, i.e., programs that carry out tasks autonomously. The agent may use only the tools explicitly assigned to it, for example reading a calendar but not triggering payments. Without this limitation, every mistake by the model would immediately become an expensive problem.

Beyond the technology sector, the term also appears in politics and economics. There are allowlists of safe countries of origin in asylum law, allowlists of permitted food additives, and allowlists of investors permitted to purchase a stock. The principle is the same everywhere: only what is listed is permitted. Anyone reading such a report should always ask two questions. Who decides on inclusion, and what happens to everything that is missing?