Al Baker, co-founder of Prose Intelligence, joins us for an episode of From the Source, the Blackdot podcast. He discusses the latest developments in artificial intelligence (AI) and their implications for OSINT and security. He also explores the limitations of current AI models, the risks of misinformation and the value of rigorous research practices in intelligence work. You can read his insights below or listen to the full episode here.
OpenClaw is a new open-source tool that, according to Baker, enables LLMs to ‘plug’ into one another to unlock new capabilities.
‘Anything that a human can do on a computer, an LLM can do on a computer via OpenClaw,’ explains Baker.
He adds that a consequence of this is enormous security vulnerabilities.
‘It's a very dangerous piece of software to use recklessly, but the counterpart to that is it’s also potentially tremendously powerful,’ he says. ‘The robots are not here to take your boring admin away just yet. But it genuinely is a matter of when and not if. The fact that the plumbing is there now and the fact that it works in principle are genuinely very exciting.’
Despite being excited about OpenClaw, Baker is sceptical that the rise of agentic AI will be smooth. He adds that it’s a good reminder that Large Language Models (LLMs) are not the only kind of AI and, in fact, aren’t the most natural ones to deploy.
‘They're a text predictor,’ he explains. ‘It's a very, very sophisticated autocorrect and that can do a lot of things, especially in the context of admin and kind of boring administrative work where mediocrity is kind of what you're going for.’
This raised a key question: to what extent can OSINTers trust AI and its output?
For Baker, it’s essential to define what ‘agency’ is when exploring AI, particularly agentic AI. He explains that an interacting network of complex, non-agentic objects doesn’t necessarily create agency.
‘If I'm playing an online game with a bunch of other people, these are non-agentic devices interacting and creating a game, but agency doesn't exist in the game. It exists in the people playing the game,’ he explains.
He believes there is some more work to do before we can say that AI, as we understand it, is agentic because, ultimately, it can’t make decisions.
‘The missing piece, and what I think is often overlooked, is trust. It's certainly not possible to trust an LLM in the way that it's possible to trust a human analyst, and that's because an AI essentially has no skin in the game,’ Baker explains.
While humans may get frustrated with AI for its mistakes, misinformation or errors, Baker explains that this is ‘pointless’, because we cannot hold the technology accountable.
‘It's about as useful as me getting annoyed at my alarm clock for forgetting to plug it in. The alarm clock doesn't have any agency in not waking me up,’ he says.
While Baker considers users' understanding of AI’s lack of agency crucial, he is passionate about the power of the right tools and technology to support OSINT investigations.
Baker and his friend Jordan founded Prose Intelligence a couple of years ago. They identified a market gap for a tool that could archive Telegram data, specifically for intelligence use cases. The tool is based on a technology that Jordan developed during his career as a journalist and later as a disinformation researcher, called Telepathy, which archives Telegram data. While Telegram is their flagship data source, Prose Intelligence has also expanded into some fringe social media platforms, smaller platforms that have recently cropped up, and some non-Western platforms.
When developing Prose Intelligence, Baker and Jordan wanted to take OSINT analysts' needs seriously, because they are very different to those of a marketing or business analyst. This includes ensuring the right data for open source intelligence use cases and providing compliant data that’s rich enough to support analytics tools like Blackdot, which specialises in network mapping.
‘One of the reasons network mapping isn’t exploited more is that the raw data, which is provided to these platforms, isn't rich enough to support it,’ he begins. ‘There are all these fantastic analytics capabilities that people are building, but the missing piece is data that is reliable enough, rich enough, and that is sourced responsibly enough.’
Baker works in misinformation but has a background in academic Philosophy before commencing his OSINT career. After completing his postdoctoral work, he was hired by Logically as Head of Fact Checking, helping position the company as a third-party fact-checker for Meta and other social media platforms.
‘What I did have from my training as a philosopher was a really solid grounding in argumentation, logic and epistemology. So, I had a better idea than most of what could reasonably be said about what is true, false, misleading, deceptive, what have you, and about good and bad ways of arguing for the truth, falsehood, or otherwise, of a particular claim,’ he explains.
For Baker, the standard definitions of misinformation and disinformation are not well suited to the actual work involved in understanding and responding to online misinformation.
“The [standard] difference between the two is that misinformation is a falsehood spread recklessly,’ he begins. ‘Disinformation is falsehood spread intentionally.’
While it may seem like a reasonable distinction, it becomes complicated when applied to the internet and social media.
‘If a state-backed actor deliberately spreads falsehoods to create disruption, that's plausibly a matter of national security,’ he explains. ‘If it's misinformation, then it's not necessarily, or it's very arguably, not a matter for national security.’
It’s a case of analysing the source and understanding the dichotomy, rather than just the intent, of any given social media post.
After 10 years in the industry, Baker doesn’t think anyone yet understands the phenomenon of misinformation, much less how to address it, despite the fantastic academic work happening.
Baker explores how technology, such as AI, has influenced the information environment. He explains that the rise of misinformation came with the rise of social media, not because people now had a platform to intentionally spread lies, but because social media became the platform for almost all communication.
‘The issue is that we had an information environment with one huge set of problems,’ he explains. ‘The old bastions of the media owned the news agenda. They got to define what was important and worth talking about… And then we replaced that information environment with social media and it solved a lot of those problems, but it also creates a whole bunch of new ones because you get rid of all of the institutional safeguards.’
When it comes to solving what Baker describes as the epistemological crisis, or ‘the misinformation problem’, he believes OSINT has a lot to offer.
‘Give a roomful of OSINT analysts, who've been working for a couple of years, a laptop each and a web browser, and they will solve incredible problems for you,’ begins Baker. ‘But you've got to define the problems that you want them to solve and give them the frameworks to provide the solution.’
He adds that the OSINT community has radically outpaced the institutions it should be serving, as well as the frameworks OSINTers can use to understand the problems it can solve. Similarly, he describes how OSINTers have come from the ‘wild west’ of the internet, often conducting ‘piratical, punky investigations’. While this has led to the interesting, welcoming, and useful community that exists today, it also often results in a lack of documentation, including methodology documentation.
For OSINTers working with the government and the private sector, documenting methodological practices is critical.
‘It is much easier for you to spend money if you know that, if asked, your supplier will produce a document that says, this is the methodology we use,’ he adds.
If OSINT has a role to play in today’s information environment, it starts with sharper questions, clearer frameworks and documented methods that others can trust.
Listen to AI, trust, misinformation and OSINT with Al Baker, and stay tuned for more OSINT insight on Blackdot’s next podcast episode.