Bounty Hunters
Have enough conversations and key themes emerge. Lately it’s all about data (or content or editorial or anything) that feeds the machines and how the machines are doing in their respective races. Open Evidence prevails in a comparison with OpenAI; in another study, it ranks over UptoDate. Every day the hunt for new and original sources continues, and the comparisons emerge.
Big tech LLMs relentlessly look for an edge. AI native companies always knew it was about breadth and depth of vertical content. And for both, the race to license or acquire differentiated assets is on. Full throttle for anything that comes with machine-ready quality, uniqueness, provenance, and rights. (See this week’s latest Outsell analysis: Building Scalable AI Licensing Strategies.)
Any self-respecting information provider has long known that garbage in means hallucinations out. Firms can’t simply hoover up the open web, have more or less the same content corpus, and expect differentiated output. It becomes one big ocean of water. No matter where one falls in, the waters are deep, choppy, salty, with no shore in sight. Bob around long enough in that ocean and it’s cold, dark, and lonely. There is only one way out or down.
It started with the likes of Associated Press or Getty. Axel Springer, Wiley and others joined in. We’d then get calls. One big tech player wanted help identifying anything with images (think cartoons or chemical diagrams) that came from highly specialized sources where provenance and rights came with the corpus.
Then the battle over 23andMe and its data corpus potentially ending up with Regeneron. The uproar could be heard around the world. Last month, Google won a bid in bankruptcy court to acquire the data assets of Spirit Airlines. 600 million email and chat records, 17 million OneDrive files, 20 million SharePoint items, 30 million recorded customer service calls, and untold volumes of flight data, fuel slips, and purchase records for over 3/4 of a million parts. You get the picture.
Employee rights? Customer call-ins? You know the kind that says this call may be recorded for training purposes? I don’t think that was ever meant to mean AI training purposes and certainly not at Google. May the flight attendants and their union win.
Salesforce and Anthropic’s Claudeforce launching just in time for Dreamforce is another harbinger of what’s to come. Sure, we could argue getting anything out of Salesforce and its circa-90s UI is a bear. Any software needing an army of people with the word ops in their job title is an anachronism. But this is about so much more than Claude being the UI. It’s about the data. BIG-DATA. Let’s hope we can avoid another HubSplot moment. We are racing to read our Salesforce contracts.
And all indications are that this little big deal will come with some semblance of transactional pricing. Are you kidding me? Let’s hope the CFOs of the world revolt!
No data set or requests for them seem to be too small. Just this morning I received an email with the subject line: your company’s data. (No one uses initial caps anymore; they have gone the way of vowels in high-tech names.)
Be that as it may, Paige from the outcome desk, who carries the title partnerships and likes to follow the tech-bro pattern of starting emails to strangers with Hey (initial cap here tho) sends the following:
Hey Anthea, we pay businesses $150k-$1M for old internal data, like sales processes, support tickets, SOPs and team messages to train AI models. Never customer or sensitive data, fully anonymized. We’ve paid businesses more than $20M in the last three months. Is this something that would be interesting for you?
I’m sure it’s an agent, spam, phishing. Paige doesn’t exist on LinkedIn; her company doesn’t either. Nor is there any semblance of her with a Google search. If you find her, let me know. But she is on the hunt for Outsell’s 30-year treasure trove and flaunting the notion of extra zeros to go along with it. Little does she know my affinity for that term.
Just like we have bug bounties, security hack bounties, and criminal bounties, fake Paiges notwithstanding, bounty hunters for data are the next new thing. And creative methods of finding data – not just in bankrupt businesses – will be rewarded. There are hundreds of new treasure troves sure to emerge, mine, and pick to the bone. Human creativity has no limits. Just ask AI.
When you need to deepen or develop your AI licensing strategy, contact us.