Anthropic legal AI launch coincides with $285 billion US market drop

Anthropic legal AI launch coincides with $285 billion US market drop

A legal software announcement shouldn't move hundreds of billions in the stock market. Yet, the unveiling of an AI legal agent coincided with a shakeup that wiped out $285 billion from the US market, according to figures circulating in financial debates. This number, repeated with irony in trading circles, summarizes the tension of a market that has been looking for any excuse to correct for months. The official trigger was a product: a model capable of automating legal and administrative tasks. The real trigger, according to the colder analysis, is another: a 66% rise over five years that was screaming for a pause.

What is the legal agent and why its impact is exaggerated

The flagship product is an agent oriented toward legal and administrative tasks. Among its functions is the triage of non-disclosure agreements, a utility that, according to critics, adds nothing that couldn't be done before with other tools. The discussion about its true scope is far from closed: some argue it is a minor advancement, a model adjustment with more parameters and a commercial wrapper, while others defend it as the spearhead of a deep transformation in the legal sector.

What is fruta is the effect on the narrative. The presentation served as a perfect excuse for part of the market to sell. Palantir and Rocket Lab fell 13-14% in the same session, two companies that have little or nothing to do with legal software. The correlation, rather than causality, points to a known pattern: when the market is overextended, any headline serves to take profits.

The technical correction no one wants to acknowledge

The most repeated analysis among those who trinc the market daily is that the drop responds to a technical correction after an excessive rally. The US market had accumulated a 66% appreciation over five years, and any excuse works to trim positions. The comparison with the first quarter of 2025, when tariffs served as the trigger for a similar shakeup, is repeated: the cycle repeats because investor psychology doesn't change.

Against this reading, some point out that the announcement is just the spark of a structural change. The emergence of AI agents in knowledge-intensive sectors—legal, administrative, accounting—could erode entire business models. The debate over whether this will happen in months or decades is, today, irresolvable.

The AI debate: between intern and total replacement

The conversation quickly shifted to the real role of AI in employment. A report cited in the discussion argues that AI can handle between 11% and 20% of a person's tasks, but not replace them entirely. The most repeated analogy is that of the intern: useful, fast, but in need of constant supervision. Microsoft's CEO reportedly backed this thesis, nuancing the narrative of total substitution.

In contrast, the most tech-optimistic current argues that the qualitative leap is yet to come. AI won't just train on internet data, but on data generated by daily work itself, which would multiply its learning capacity. The prediction that AI capability doubles every six months is used as an argument to justify that the impact will be exponential, not linear.

The skeptical stance reminds us that current models don't understand the world: they predict the next token or state based on their training. Without real understanding, complete substitution is a fantasy. The debate over whether AI learns or merely appears to remains open.

The coming labor market: retraining or scrapping?

Beyond the stock market, the debate delved into the future of work. Some participants argue that AI will reconfigure entire sectors, from law to programming, and that the key will be adaptability. Elon Musk's quote about there being no code, but neural networks generating pixels and commands, was used as an example of the approaching horizon.

Another current focuses on the distribution of benefits. If AI frees up human time, the question is whether that time will be dedicated to useful tasks—care, maintenance, essential services—or to swelling unemployment and the informal economy. The warning that those who don't keep up with technology will be left out of the labor market intersects with criticism of a system that prioritizes speculation over productive work.

Real use of AI: reports, analysis, and workflows

Beyond predictions, there are concrete experiences. One participant describes how they use AI to generate reports in perfect English, correcting errors afterward: what used to take a week now gets done in a day. Three perfect analyses in one day is the type of data that illustrates the impact on individual productivity.

However, the same person warns that AI doesn't analyze: it automates. To get value from it, you must analyze workflows, identify bottlenecks, and decide how to integrate it. The fantasy of total replacement clashes with the reality that, today, agent flows in real business processes are still scarce. Proof of concepts abound; real implementations, less so.

Skepticism: bubble or revolution?

The shadow of the dot-com bubble looms over the conversation. Some compare the current enthusiasm with that of the late nineties and recall that the internet ultimately became 80% porn and social media. The feeling of a familiar smell of overvaluation is repeated. AI will have utility, as the internet did, but that doesn't guarantee that current prices are justified.

At the other extreme, the most optimistic current argues that this is different. AI doesn't depend on human capacity, which is finite, but on its own capacity, which multiplies. The arrival of quantum computing in coming decades would add another layer of disruption. The discussion, like almost all touching the future, resolves with an we'll see.

How much of the $285 billion drop is due to the legal agent and how much to the desire to sell? The answer, today, lies with the market. And the market, as always, gives no explanations.