Evidence since Jack Dorsey’s resignation indicates that Twitter/X’s overall approach to speech has become less restrictive, not more.

Key points:

1. Major rollback of COVID-19 moderation rules. Under Dorsey, Twitter removed ~97,000 pieces of COVID-19 misinformation and suspended over 11,000 accounts from 2020–Sept 2022 under a strict health-misinformation policy. After Elon Musk took over, Twitter announced on Nov 29, 2022 that it stopped enforcing this policy and would no longer label, demote, or remove COVID falsehoods. (time.com) This is a clear move toward less restrictive moderation.

2. Loosening protections against anti‑trans harassment. In early April 2023, Twitter quietly removed an explicit rule that prohibited targeted misgendering or deadnaming of transgender people—an anti‑abuse protection that had existed since 2018. (techcrunch.com) X later partially reinstated a narrower rule in 2024 that focuses mainly on repeated, targeted harassment and often just reduces visibility of offending posts rather than reliably removing them. (them.us) Net effect: weaker, not stronger, speech restrictions in this area.

3. Dissolution and downsizing of trust & safety infrastructure. After Musk’s acquisition, Twitter/X dissolved its Trust and Safety Council and leaned more heavily on automation while hate speech surged, a shift widely described as relaxing moderation. (en.wikipedia.org) Regulatory filings from Australia’s eSafety Commissioner and related reporting show X:

   • cut global trust & safety staff by ~30%, including 80% of safety engineers;
   • halved the full‑time content moderation team and significantly reduced contracted moderators;
   • became slower and less responsive to reports of hateful content. (apnews.com) These cuts reduce the platform’s actual capacity to restrict content compared to the Dorsey era.

4. Mass reinstatement of previously banned accounts. Musk explicitly pursued a “general amnesty” for suspended users, restoring accounts banned for COVID misinformation and other policy violations, and reinstating high‑profile accounts such as Donald Trump and Marjorie Taylor Greene. (cnbc.com) A wave of reversals of earlier enforcement decisions is inconsistent with a move toward more censorship.

5. Empirical studies describe moderation relaxation and more visible hate content. Academic work and monitoring organizations examining the post‑acquisition period conclude that Musk’s takeover was associated with relaxed moderation and substantial increases in hate speech on the platform (including anti‑LGBTQ and racist content). (montclair.edu) While increased hate speech doesn’t by itself prove there are fewer takedowns, these studies explicitly tie the surge to moderation being loosened, not tightened.

There is some evidence that X has been more willing to comply with certain state censorship demands (for example, account withholding by jurisdiction). (arxiv.org) But the prediction was broadly about Twitter’s own content‑moderation and speech restrictions being ramped up relative to Dorsey’s tenure. Given the large‑scale rollbacks of major policies, staff and structural cuts, and mass reinstatements, the dominant trend since late 2021 has been less restrictive speech moderation overall, so Sacks’s prediction that “things are only going to get worse” in the sense of more censorship is best judged as wrong.