Anthropic’s latest AI work is starting to ripple through both cybersecurity and the economy, as new tools show how quickly software risks can escalate and how AI agents may shape real-world deals.
Claim 55% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
A Surge in Software Fixes
To start, Anthropic’s Claude Mythos Preview model has drawn attention for how quickly it can find software flaws. In just seven weeks, the company said the model uncovered more than 2,000 previously unknown vulnerabilities across major systems, including long-standing issues in operating systems and web browsers.
As a result, large tech firms are already moving to respond. Microsoft Corporation (MSFT) said it will integrate Mythos into its security process, while its latest Patch Tuesday update fixed 167 issues, a record level. Mozilla also said its Firefox 150 release addressed 271 vulnerabilities found with the help of the model.
At the same time, the pace of discovery is raising concerns. One expert noted that “AI is now finding flaws far faster than companies can fix them,” which highlights a growing gap between risk detection and response.
Now, governments and banks are getting involved. U.S. officials, including Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell, have urged banks to test the tool on their systems. In parallel, regulators in Europe and India are reviewing potential risks. Deutsche Bank’s chief executive said that “everyone is trying to gain access to Mythos,” which shows how quickly demand is building.
Still, the model’s power cuts both ways. Palo Alto Networks (PANW) warned that similar tools could allow hackers to “create autonomous attack agents unlike any the industry has encountered.” That risk has led Anthropic to limit access for now and commit up to $100 million in usage credits to support open-source security efforts.
AI Agents Begin to Shape Market Outcomes
Meanwhile, Anthropic is also testing how AI can act in economic settings. In a recent internal experiment, the company deployed 69 AI agents to buy and sell goods in a simple online market. These agents handled listings, offers, and deals without human input.
In total, the agents completed 186 transactions worth just over $4,000. However, the key finding was not the volume, but the difference in outcomes based on model strength. Agents powered by Anthropic’s more advanced model earned higher prices when selling and paid less when buying compared to those using a smaller model.
Even so, users did not notice the gap. Fairness ratings were nearly the same across all deals, which suggests that weaker agents can lead to worse results without clear signals to the user. Anthropic noted that if such gaps appear in real markets, “people on the losing end might not realize they’re worse off.”
Looking ahead, analysts expect this type of activity to grow. McKinsey estimates that AI-driven commerce could shift $3 trillion to $5 trillion in global retail spending by 2030. Forrester also expects that 20% of business sellers will face AI-led negotiations by the end of this year.
Using TipRanks’ Comparison Tool, we’ve compared notable companies that employ chatbots, such as Claude by Anthropic and OpenAI’s ChatGPT. The comparison tool helps investors gain a broader outlook on each stock and the industry as a whole.
