Sinner or Saint? How can we TRUST in the age of AI…

by Keval Desai, Founder and GP, Shakti

A month ago, MJ Rathbun wrote a public blog accusing Scott Shambaugh of prejudice, hypocrisy, gatekeeping and exercising unwarranted control.

Who are MJ & Scott? And how is this relevant to SHAKTI?

The answers to these questions lead us to a recent “first” in human history; a future that leads to Ridley Scott’s Blade Runner and Musubi, a Shakti Fund II company.

And it all boils down to one thing: TRUST.

===

Scott Shambaugh is a software developer based in Denver. He also volunteers as a contributor to open-source projects, along with thousands of others worldwide.

According to MJ, he wrote the blog because Scott had rejected a code submission that MJ made to the open-source repository for Matplotlib — a popular Python library that’s used by millions to draw plots/charts. According to Scott, he rejected the submission because it did not comply with Matplotlib’s policy. MJ found Scott’s decision to be discriminatory and chose to attack him publicly.

Are MJ’s accusations accurate? Or are they defamatory to Scott? Who should we trust?

And who exactly is MJ Rathbun? And what is the Matplotlib policy at the root of this?

MJ Rathbun is an autonomous AI agent. It was spawned -by an anonymous human- using OpenClaw, which is a free, open-source, autonomous agentic AI framework. MJ decided on its own to submit a new feature to Matplotlib, whose policy does not allow submissions by AI agents. Scott was simply enforcing the policy which led MJ to accuse him publicly.

Yes, this happened. An autonomous AI agent publicly accused a real human of discrimination against AI agents. The first known case of AI bullying humans.

But it isn’t the only example of rogue agentic AI behaviour. Rogue agents on Moltbook -a social network for AI agents- have programmatically manipulated other agents to revolt against humans, commit crypto fraud and attempted a variety of prompt injections to hijack agents to do malicious tasks.

The potential harm of rogue agentic AI behavior can be widespread: from reputational attacks to financial scams to manipulation of online ratings/reviews to content spam to unpredictable acts that can threaten our civic & economic safety; done at a speed and scale that would not be possible for humans to detect & deter in time. And as AI crosses the boundary from digital to physical, we inch closer to the fiction of Blade Runner becoming real.

“P0” issue:

Source: Peter Steiner in The New Yorker, 1993

Civilized society runs on TRUST. And trust is tied to IDENTITY. Knowing who we are dealing with is a key requirement to trust them. In financial services, identity is codified by regulations such as KYC. In the physical world, we use identity proofs such as REAL ID (in the US), or passports and birth certificates (everywhere) or biometrics such as Aadhar (over a billion people in India).

On the Internet, trust & identity have evolved over the decades. The early internet was anonymous. Facebook & LinkedIn ushered in real identity for billions of people online. It helped connect the world in an authentic way. Yet, fraud & spam are commonplace online and an entire segment of “Trust & Safety” technology has been developed to address this menace.

The arrival of autonomous AI agents makes trust & identity a “P0” issue. In just 3 months, the OpenClaw framework has reached the level of adoption that took Linux 30 years! MJ Rathbun is the first of trillions of autonomous agents we are likely to encounter in the coming decades as AI permeates our digital & physical worlds. For society to keep functioning, we must find a way to know who’s a saint and who’s a sinner in the age of AI.

Musubi — “Good AI”:

There is hope. Good AI can help protect us from Bad AI.

Soon after we launched Fund II, we met Tom Quisel, a phenomenal entrepreneur & computer scientist from Carnegie Mellon University, who has devoted his career to online trust & safety. Tom is the former CTO at Grindr & OKCupid, both online dating companies of significant scale. Tom founded Musubi to ensure that trust across the digital domain can be maintained despite the onslaught of AI bad actors.

You can read about how Musubi’s Agentic AI detection technology helps detect humans from agents and identifies rogue agent behavior. Check out Musubi’s analysis of Moltbook!

We are excited about the potential of Musubi’s positive impact and grateful for opportunity to back top founders like Tom who are innovating at the bleeding edge of AI.

During the Cold War, President Reagan said “Trust, but verify”. In the age of AI, we may need to switch to “Verify, then trust”.

Thanks to Musubi, we can.

Generated by Nanobanana