Musk immediately terminated twitter's general counsel and head of legal policy, trust, & safety - this does not bode well with anticipated 3/4 downsizing:
>2. A stripped-down trust-and-safety team is unable to deal with government subpoenas or complex law-enforcement requests. A bare-bones team might, for example, accidentally assist outside efforts to identify anonymous dissidents and activists in foreign countries.
>3. The trust-and-safety team is unable to stop coordinated efforts from fraudsters orchestrating low-level scams. Similarly, a strapped trust-and-safety department is unable to combat or monitor child-sexual-abuse material, sex-trafficking efforts, nonconsensual pornography, and copyright violations.
…
>One former trust-and-safety engineer for a large social network told me that many elements of the job that seem boring or straightforward are actually incredibly fraught, like how to define and take action on different kinds of spam. Trust-and-safety officers in charge of such efforts aren’t just dealing with Viagra ads or crypto-scam bots; they’re figuring out how to handle bulk messages from legitimate political organizers exploiting the platforms for mass messaging. As one person put it, there are good actors and bad actors and also “spammy but not necessarily malicious businesses trying to get you to buy things in between, and all those things can look very similar to machine-learning models.”
>Those with trust-and-safety experience at the platform told me that a big percentage of the job is dealing with the messy edge cases that are difficult for a computer to decipher. Programs might be able to address specific product quirks if a user files a clear help ticket reporting an obvious problem. “But if I wrote in, ‘My account has been hacked because it “accidentally” liked a porn tweet on 9/11 and I’m U.S. Senator Ted Cruz,’ that’s going to be a lot for a computer to unpack,” Brian Truebe, a former Twitter trust-and-safety professional, told me over email.
>“A lot of things humans say and do are only easily interpretable/decoded by other humans,” he continued. “And when all speech is happening in a few places, those few places need more humans to review, not fewer.”
>Reactionary tech figures such as Musk like to imply that content-moderation teams act as a kind of thought police. But these teams largely work on protecting users’ privacy, complying with laws, or keeping the site from becoming overrun by the kind of spam that no human wants to encounter. “To really have a robust security-and-abuse team, you need a massive amount of actual humans to respond and filter things that need to be filtered out,” Southey Blanton, a systems technician who worked in trust and safety at MySpace, told me. Blanton said that cuts to his team led to a skeleton crew of moderators, who had to rely on imprecise AI tools to get rid of bots and spam—which led to many legitimate human accounts getting banned as well. “Overall, a social-media site is under attack, as well as being overwhelmed, basically 24/7, 365,” he said. “I am fully convinced that if Musk does what he is saying he will do, it will be an absolute shitshow.”
https://www.theatlantic.com/technology/archive/2022/10/elon-...