Article · 16 October 2026

AI filters, the OpenAI lawsuit and Anthropic's growth: what each piece says

Security, money and law in a single day. Three lines that forward themselves, and what each linked piece really says.

Alejandro Carmona · Nopubly · 16 October 2026

The line that will circulate today fits on one row: terrorists can get around AI filters. It is almost word for word the headline in ABC: the hidden danger of AI, terrorists can bypass its safety filters to get the information they need. It sounds like a done deal. It is worth checking what the piece actually promises.

ABC opens with an image: artificial intelligence arrived as an open door to infinite knowledge, and that hides a security gap for the whole of society. It adds that chatbots have barriers designed to... and there the text we have is cut off. The verb in the headline is 'can'. It describes a possibility. It does not announce an attack or a specific case. Method, examples and sources, if there are any, are in ABC's piece and not in the forward.

The number that sounds like a win

The second line travels with a figure: Anthropic grew almost 50%. Expansión headlines that the company has told investors its quarterly revenue rose almost 50%. Pasted alone, the figure looks like a victory. The lead says something else: that rapid growth is below what some players expected, at a time when worry about AI is unsettling markets. Same number, two readings. The screenshot says 'growing fast'. The piece says 'growing less than some hoped'.

The lawsuit that sounds like a verdict

The third is the most legal. ABC reports that the attorneys general of eight US states have sued OpenAI. The headline puts the accusations in quotation marks: 'causing harm to children', 'deceiving parents' and breaking the law. The lead places the suit on Wednesday and stresses that two of those states are governed by the Republican Party. A lawsuit is an accusation. It is not a conviction. And those quotes belong to the people suing, not to proven facts. A forward saying 'OpenAI guilty of harming children' skips an entire court.

Together, the three make a full day. One says the tool can be bent. Another says the money is growing, but less than expected. The third says some prosecutors want someone to answer for it. Security, money and law. Nothing in the headlines links them, and we should not pretend otherwise: the thread is ours, not theirs.

Un análisis más profundo de las implicaciones

What follows is our reading, not data from the pieces. These are predictions, and we offer them as such.

First, the filter will move from a technical problem to a legal one. Attorneys general in eight states are talking about children and misled parents, and ABC warns that the barriers can be bypassed. The underlying question is who answers when a barrier fails. We expect more suits of this kind and more pressure on companies to show what they do, not just say it.

Second, the market will price that risk. That growth near 50% can be read as lower than expected says a lot about how high the bar has been set. When the bar is that high, every new cost weighs. Legal defence, safety teams and audits become a line item, not a footnote. We expect investors to ask about them at every earnings round.

Third, work. If safety and compliance become enforceable, demand grows for profiles that are rare today: people who test models looking for failures, people who write usage policies, people who review what a chatbot tells a minor. This is not mass employment. It is a niche that appears where there used to be an 'accept' button.

One idea for the reader. AI was sold as an open door. Increasingly it will be sold as a door with a lock, someone responsible and an invoice. Whoever understands that shift before the forward does will be ahead.

Qué hacer antes de reenviar

• With the terrorist line: read ABC's whole piece. 'Can' is not 'have'. Do not pass it on as if there were a specific case.
• With the 50%: add the second half of the sentence. According to Expansión, it is less than some players expected.
• With the lawsuit: write 'sued', not 'guilty'. These are accusations from the attorneys general of eight states, according to ABC.
• If you have not opened the link, do not forward it. If you have, name the outlet.

Questions

Does ABC say an attack was planned with AI?

Not in the headline or the lead. They say terrorists 'can' get around chatbot safety filters. That is a possibility, not a specific case.

Is Anthropic growing or disappointing?

Both, according to Expansión. Quarterly revenue rose almost 50%, but the growth is below what some players expected.

Has OpenAI been found guilty?

No. According to ABC, the attorneys general of eight US states have sued it. That is an accusation, not a verdict.