Top Meta staff allegedly compared Instagram to a drug and worked for years to obscure the social media platform’s potential dangers from parents and children, even as they appeared to acknowledge their technology was harmful, according to newly unsealed internal communications cited in a court filing.
See also Reuter's reporting on the lawsuit:
Meta shut down internal research into the mental health effects of Facebook after finding causal evidence that its products harmed users’ mental health, according to unredacted filings in a lawsuit by U.S. school districts against Meta and other social media platforms.
[...]
The internal documents cited by the plaintiffs allege:
Meta intentionally designed its youth safety features to be ineffective and rarely used, and blocked testing of safety features that it feared might be harmful to growth.
Meta required users to be caught 17 times attempting to traffic people for sex before it would remove them from its platform, which a document described as “a very, very, very high strike threshold."
Meta recognized that optimizing its products to increase teen engagement resulted in serving them more harmful content, but did so anyway.
Meta stalled internal efforts to prevent child predators from contacting minors for years due to growth concerns, and pressured safety staff to circulate arguments justifying its decision not to act.
[...]
In a text message in 2021, Mark Zuckerberg said that he wouldn’t say that child safety was his top concern “when I have a number of other areas I’m more focused on like building the metaverse.” Zuckerberg also shot down or ignored requests by Nick Clegg, Meta's then-head of global public policy, to better fund child safety work.
I might not want to live in Meta's metaverse...
Meta internally projected late last year that it would earn about 10% of its overall annual revenue – or $16 billion – from running advertising for scams and banned goods, internal company documents show.
Solutions include charging known scammers higher prices, leading to more profit for Meta.
This is X in 2025: Potentially fake accounts crying at other potentially fake accounts that they aren’t real, all while refusing to acknowledge that they themselves aren’t who they say they are — a Russian nesting doll of bullshit.
X's own features say foreign nationals are posing as Americans to drive political engagement, mis-, and disinformation.
An advisor to al-Qaida. One of the founders of Hezbollah. The head of an Iraqi militia group known for attacks on U.S. troops. And a top official with the Houthi rebels who recently lashed out at the “criminal Trump.”
These are among the U.S.-sanctioned terrorists who appear to have paid, premium accounts on Elon Musk's X, a new Tech Transparency Project investigation has found, raising questions about the platform's dealings with individuals who have been deemed a threat to U.S. national security.
The new Fairwater AI datacenter in Wisconsin stands as a remarkable feat of engineering, covering 315 acres and housing three massive buildings with a combined 1.2 million square feet under roofs. Constructing this facility required 46.6 miles of deep foundation piles, 26.5 million pounds of structural steel, 120 miles of medium-voltage underground cable and 72.6 miles of mechanical piping.
Unlike typical cloud datacenters, which are optimized to run many smaller, independent workloads such as hosting websites, email or business applications, this datacenter is built to work as one massive AI supercomputer using a single flat networking interconnecting hundreds of thousands of the latest NVIDIA GPUs. In fact, it will deliver 10X the performance of the world’s fastest supercomputer today, enabling AI training and inference workloads at a level never before seen.
Taking inspiration from a rare mutation that makes people impervious to viral diseases, a Columbia researcher is developing a therapy that could bestow this superpower on the rest of us.
Today, we are excited to introduce an autonomous AI agent that can analyze and classify software without assistance, a step forward in cybersecurity and malware detection. The prototype, Project Ire, automates what is considered the gold standard in malware classification: fully reverse engineering a software file without any clues about its origin or purpose. It uses decompilers and other tools, reviews their output, and determines whether the software is malicious or benign.
Instead of saying "we cannot ensure that no harmful sequences will be generated by our function, partially because we don't know how to specify and enumerate harmful sequences", we talk about "behaviors", "ethical constraints", and "harmful actions in pursuit of their goals". All of these are anthropocentric concepts that - in my mind - do not apply to functions or other mathematical objects. And using them muddles the discussion, and our thinking about what we're doing when we create, analyze, deploy and monitor LLMs.
There’s a lot of excitement about what AI (specifically the latest wave of LLM-anchored AI) can do, and how AI-first companies are different from the prior generations of companies. There are a lot of important and real opportunities at hand, but I find that many of these conversations occur at such an abstract altitude that they border on meaningless. Sort of like saying that your company could be much better if you merely adopted _more software_. That’s certainly true, but it’s not a particularly helpful claim.
This post is an attempt to concisely summarize how AI agents work, apply that summary to a handful of real-world use cases for AI, and to generally make the case that agents are a multiplier on the quality of your software and system design.
Of the many explainers I've seen and read, this is probably one of the best.