Security expert Bruce Schneier has a rosier outlook on AI than most, arguing at the RSA Conference this week that the technology will “be one of humanity’s most important inventions,” and may one day be used to write legislation and—eventually—eliminate politicians.
You can’t attend RSA without encountering Schneier; his expertise runs the gamut, and a quarter-million people follow his Schneier on Security blog. His presentation started by arguing that “when you replace humans with AI, you see differences in speed, scale, scope, and sophistication.”
AI can be an excellent summarizer of existing information. It’s good at explaining and can be a teacher with infinite patience, he said. It can predict with some accuracy based on its training, and it has the potential to be good at assessing. Eventually, it can manage decision-making.
“I’m gonna be largely optimistic and not dwell on details,” said Schneier, “though I’m less optimistic about the social implications. What needs to happen before people trust AI? What could go wrong? What can we as security techs do to help?”
AI and Democracy
“Democracy is a system for distributing decisions across a population,” noted Schneier. He explained that AI can inform five areas: politicians, lawmaking, administration (including bureaucracy), the legal system, and even citizens.
How could the use of AI by politicians go well? Of course, AI can help with persuasion. Pols could create personalized chatbots, help fundraise, and conduct polls.
![](https://sm.pcmag.com/pcmag_me/photo/s/security-e/security-expert-bruce-schneier-at-the-podium_68ac.jpg)
“There could be an arms race as politicians start using these tools,” said Schneier. “We don’t know if the technology would favor one side or another.”
He pointed out that any major politician today is just the public face of a complex system. AI changes the nature of politics, but so did the advent of television.
As for legislation, AI can summarize letters, comments, and meetings, and make sense of constituent input. It can also assist in lobbying strategies and even make laws. Schneier noted that in December the city of Porto Alegre, Brazil, enacted a law about water meters that was written by ChatGPT.
“A law is just a piece of generated text that a government agrees to adopt,” said Schneier. “AI can take human-written laws and explain what they really mean. It’s good at finding or creating legal loopholes, similar to finding vulnerabilities in software. AI can also help identify unintended consequences and determine how laws interact with each other.
“Going forward, AIs can write laws that are impossible for humans to understand,” said Schneier. “Suppose AI gets really good at identifying reckless driving. We could pass a law making the AI’s judgment official.”
In the realm of administration and bureaucracy, AI won’t get bored and could step in to assist with benefits administration; there are too few humans to do that. “Could an AI come up with better institutional designs?” said Schneier. “Would we implement them?”
Meet Your AI Public Defender
“AIs can be lawyers, though early attempts did not go well,” said Schneier. “Chatbots can now cite real sources and minimize errors. Future AIs will be better at legalese.”
He pointed out that an AI public defender might be more effective than an overworked pro bono lawyer. “Even then, the rich will get a human lawyer plus AI while the poor will just get the AI.”
Schneier noted that, at present, being sued is a big deal, a signal that someone is willing to spend a lot of money challenging you. If AI reduces the cost of a lawsuit to nothing, that signal will be lost. We’ll have to deal with orders of magnitudes more lawsuits.
Schneier sees a place for AI in law enforcement. AI can auto-identify tax cheats or flag fraudulent government applications. “We don’t have enough humans for this,” said Schneier. “AI can do it faster and at a bigger scale. The problem comes when the court believes the computer is always right.” He pointed out that even now, you can’t contest the result of a breathalyzer test, just the calibration and the performance of the test.
“AIs can perform judging tasks, weighing evidence, and making decisions,” said Schneier. “In many areas, we don’t have enough adjudicators. AI could offer everyone immediate justice. Think of the benefits! Maybe you’d get an AI judge first [and a] human judge for appeals.”
He went on to speculate on the possibility of binding arbitration by AI. If we decrease the cost of arbitration, would that increase the number of disputes? Will we lose the ability to resolve disagreements on our own?
Earlier this year, Supreme Court Justice John Roberts argued that “legal research may soon be unimaginable without” AI, but said it won’t replace humans and should be approached with “caution and humility.”
The AI-Empowered Citizen
One idea people may be more willing to get behind? Eliminating politicians.
Research shows that AI can already extrapolate your political preferences. An AI assistant trained on your preferences could advise you, act as your representative, and perhaps even vote on your behalf.
“Our system empowers elected officials to represent our interests,” said Schneier. “That selection process is incredibly inefficient. We all have complex wants, but we can only choose a few candidates. Your personal AI could directly participate in debates for you, along with millions of others, coming to a consensus.”
What’s Ahead?
All of these scenarios would require serious security safeguards.
“Imagine two countries using AI in trade negotiations,” Schneier said. They’ll use every means possible, including hacking the other country’s AI.
“Power matters and AI changes the balance of power. Can we make it reduce rather than increase power imbalances? The future of AI in democracy will require AI models owned by the people, not owned by monolithic corporations.
“I think AI will be one of humanity’s most important inventions,” concluded Schneier. “We actually don’t know if this is us inventing it, or if today’s systems are just overhyped. AI is coming for democracy. Let’s help tilt things to the positive.”
This post was originally published on 3rd party site mentioned in the title of this site