Op-Ed: AI forensics: How the law is building AI into daily life and law

AI may or may not be regulated to any degree. All technologies can be abused and weaponized. The scope and range of AI usage, however, has created a massive field for crime and law.

As AI spreads, the shockwaves are creating new legal environments and situations in everyday life. Not everyone’s happy about it.

Even your identity, the basis of your most fundamental rights as an individual, is now a commodity. This is where “AI forensics” come into play, and it’s a messy picture. To protect your rights, you have to know your rights and be able to get justice.

The law does have at least one leg to stand on, despite appearances. It can address these issues as basic law. AI isn’t “unregulated” or “deregulated” in this sense, and it can’t be. Any sort of loss, injury, or malfeasance can still be legally actionable.

AI forensics in practice

AI forensics are steadily growing a profile in tech news and making news in the mainstream. The definition of AI forensics includes both the forensics for AI and the forensics of AI, according to Gemini. These are types of evidence gathering, as the name suggests, but that’s just the start.

Plaintiff vs defendant isn’t and can’t be quite that straightforward, particularly in an adversarial legal environment, especially on the subject of AI. Quality of information dictates outcomes.

AI forensics also have to be accepted by courts as evidence to make a case and to be challenged. Every noun and verb in every language can be expected to testify, given the current state of litigation.

These are also the sort of legal matters that turn into instant legal precedents, and they can affect anyone using AI or anyone targeted by AI. This is the other kind of regulation, “regulation by case law”.

One successful case can generate many more similar cases. All legal cases have their own specifics, and there’s inevitably a progressive buildup of precedents which may apply to other cases.

The biggest problem is volume. AI can generate actual or potential legal issues in huge volumes. Accountability for AI behaviour is also unclear. Exactly who’s responsible for what is never simple.

Try this verbal-crayon-like depiction of the realities of legal liability for AI as a few questions and answers in a conversation:

Is Big Tech responsible for anything or everything? Not if they can help it.

How about service providers? Maybe, who knows?

AI suppliers and resellers? Anyone’s guess.

Chip makers? Forget it.

Third parties? Prove it.

We’ve moved from the “presumption of innocence” to the “presumption of evasion” in a few sentences. That’s exactly the issue, and it won’t go away until the law has a grip on the subject.

Regulation alone doesn’t solve any of these issues to any extent except in name only. Only practical law can do that.

What’s not being addressed in the regulation argument is how to make AI laws work and stick. Add to this the fact that normal legal processes must be upheld and viable for AI regulation to work at all in practice.

“Normal legal practices” are likely to be busy. Appeals are critical to the application of real law. Statutes aren’t immune to challenges. Court decisions can be overturned. This is just the laundry list for the most basic legal actions.

The evolution of AI forensics and real law

The evolution of AI forensics is a sight to see. Detecting AI has quickly moved on to identifying AI by forensics. Any sort of AI artefact or related information like servers or back-end forensics of AI processes, can pin down a perpetrator or exonerate. This information is also subject to challenges.

It’s a huge potential mass of data for and against that has to find its way into a court. Whatever information is available must be turned into comprehensible evidence to be accepted or not. See any possible issues for a court decision?

There’s also the unpredictability of AI acting on its own. AI agents are not legal entities that can sue and be sued. The theory is that the person operating the AI agents is the accountable entity, but what if that person can prove that the AI agent acted on its own initiative? It has happened.

Can you reasonably be expected to control and be responsible for the actions of an autonomous AI agent? Maybe, and it’s through your prompts. You’d need forensics to prove that your prompt was at fault or not.

You can also expect that “expert testimony” will define legal outcomes. AI experts may be expected to unravel whatever hideous situations have gone to court daily.

What AI regulation can and can’t do

There’s a tendency to assume that AI regulation is any sort of fix for AI issues.

AI regulation can:

Clarify direct liabilities.

Create rules for AI properties and behaviours backed up by compliance protocols.

Require evidence of compliance for certification and licensing, etc.

Enforce compliance by penalties and legal remedies.

AI regulation cannot:

Act in any way contrary to human rights or deprive humans of their legal rights.

Contradict, obviate, or obstruct other existing laws.

Create no-go zones for legal remedies in cases involving AI

AI forensics is the working machinery of regulation at the qubit level. When we talk about AI regulation, we’re really discussing a new form of Crimes Act. This is where AI regulation and forensics must go.