Aliya Bhatia

Twelve "child safety" bills in Congress would end online anonymity Twelve bipartisan "child safety" bills would require age verification, linking real identities to online behavior.

The curious case of pushing for legislation that will end anonymity online while running a campaign to target speech and speakers you don't like online and successfully masking it all under the guise of protecting kids online :)

boingboing.net/2026/03/06/t...

Black-box AI and cheap drones are outpacing global rules of war From 3D-printed drones to Anthropic’s Claude, advanced technologies are making conflict more accessible and less accountable — leaving human oversight at risk.

"Do we have the right rules in place and accountability norms to handle the exponential growing use of these tools? My answer would be no.”" A must-read by @rinachandran.bsky.social on the rise of dubious autonomous military tech:

restofworld.org/2026/anthrop...

ICE detains reporter Estefany Rodríguez in Nashville - Nashville Banner Reporter Estefany Rodríguez, detained by ICE with no arrest warrant, may face deportation. Her attorneys seek immediate review of the legality of her case.

ICE has arrested and detained a Nashville journalist who reported stories critical of ICE. She’s married to a U.S. citizen and has been seeking asylum here after fleeing death threats in Colombia because of her journalism there.

They’ve already sent her to Louisiana.

Overcoming Fear and Frustration with the Kids Online Safety Act KOSA will not move forward without compromise, say John Perrino and Dr. Jennifer King from Stanford University.

Congress and advocacy groups should be embarrassed to drag parents who lost their children to a partisan hearing on bad legislation that won't protect children online.

Do better! There are so many actual experts ready to help.

I wrote on this years ago with Stanford's Jennifer King.

FTC Issues COPPA Policy Statement to Incentivize the Use of Age Verification Technologies to Protect Children Online The Federal Trade Commission issued a policy statement today announcing that the Commission will not bring an enforcement action under the Children’s Online Privacy Protection Rule (COPPA Rule) aga

The FTC just announced that the only national law protecting children's data online won't apply to age verification.

You know. The Federal Trade Commission. The agency that's in charge of keeping people safe and enforcing the only rules on the books to protect children online. 🤦‍♂️

Congress’ Child Safety Bills Sound Good. Families Suggest They Won't Work. Lawmakers risk advancing bills that may not be effective nor in line with what some parents and teens actually want, Michal Luria and Aliya Bhatia write.

New piece by @mluria.bsky.social and me on what today's hearing on kids' safety is missing: the perspectives of parents and teens who are deeply skeptical of whether these bills will meet their needs. Huge thanks to @viacristiano.bsky.social for his sharp edits. www.techpolicy.press/congress-chi...

India has currently blocked at least 43,000 domains. New study shows scale for the first time A comprehensive audit of Internet Service Provider-level blocking finds far wider filtering than previously known, and it flags inconsistencies across ISPs.

Risks of sovereign control of the tech stack: Indian ISPs blocked nearly 15,000 sites including those ending with .gov.

theprint.in/science/indi...

Thanks to @knightcolumbia.org for giving me the opportunity to write about this as part of their larger project on reconstructing free expression - it’s such an important effort and I can’t wait to see what comes out of it.

What it takes to get an H-1B visa in Trump's America An immigration lawyer explains the behind-the-scenes challenges of getting the coveted U.S. visa in 2026.

"If a candidate is obese or has asthma, the officer can refuse their visa."

An immigration lawyer on what the H-1B process actually looks like in 2026: https://restofworld.org/2026/get-h-1b-visa-trump-america/?utm_campaign=row-social&utm_source=bluesky&utm_medium=social

Radiohead is demanding that the Trump administration take down a pro-ICE promotional video featuring a version of its song “Let Down” without permission.

Full statement from the band:

Looking forward to speaking at UChicago's Tech for Good conference this afternoon (1230pm CT/130pm ET) on platform governance, accountability, and users' rights. Tune in! techforgoodconference.org

Callers to Washington state hotline press 2 for Spanish and get accented AI English instead Callers to Washington state’s driver’s license agency who select automated service in Spanish are instead hearing an AI voice speaking English with a strong Spanish accent.

Govs want to use AI to translate public service information but without vetting these tools end up in embarrassing situations where their services end up no better than you lip-syncing a Bad Bunny song, or worse, in Washington state: speaking English in a Spanish accent

apnews.com/article/wash...

Read @aliyabhatia.bsky.social on the need for multilingual evaluations of AI models.

As these systems become further integrated into existing services and content moderation practices, how they work (or don't work!) in languages other than English becomes even more critical.

Finally, evaluations are just one piece of the puzzle. Rigorous testing, multistakeholder participation, and human rights-driven accountability mechanisms are critical to ensuring that users, no matter the language they speak, have access to safe, accountable, and reliable AI systems.

For the Delhi Commitments to matter, companies must ensure that they are engaging with language, subject-matter, and human rights experts and not sidelining them to ensure systems aren’t just mimicking language but actually producing safe, accurate, and helpful outputs not just mere translations

Can the bot really ‘speak your language’? It’s time India demanded multilingual AI Accountability Indian users are a rapidly growing consumer base for chatbots. Their experience should be equal, regardless of the language they speak

Multilingual evals used by major developers are often machine-translated making models over-claim multilingual capabilities and lose insights on whether systems are actually safe, reliable, and meeting users' on-the-ground needs indianexpress.com/article/opin...

Lost in Translation: Large Language Models in Non-English Content Analysis Western technology companies have long struggled with offering their services in languages other than English. A combination of political and technical challenges have impeded companies from building ...

Companies have often treated multilingual performance as an afterthought, and when they haven't, poor testing and eval regimes limit developers and deployers’ understanding of how well systems actually work in these languages, as our previous research at @cdt.org has shown: cdt.org/insights/los...

Can the bot really ‘speak your language’? It’s time India demanded multilingual AI Accountability Indian users are a rapidly growing consumer base for chatbots. Their experience should be equal, regardless of the language they speak

The Delhi Commitments launched at the #IndiaAIImpactSummit have rightfully centered multilingual evals as a core part of enhancing AI system use but how language experts will be included and eval outcomes are used is also important as Aditya Vashistha and I explain: indianexpress.com/article/opin...

India’s ‘AI Impact Summit’ Promises Little More Than Spectacle The fourth in a series of convenings that kicked off in Bletchley Park in 2023, the summit is the first held in the Global South, writes Apar Gupta.

"Any claims of AI solving for population scale problems must be matched against the reality of the Global South, in which young boys were applying enamel paint on the railings... with their bare hands to prepare for the arrival of Silicon Valley CEOs." www.techpolicy.press/indias-ai-im...

Verge headline: ICE is pushing Minneapolis underground by Gaby Del Valle Photos by Jack Califano Photo depicts a white brick wall plastered with two recurring posters: One that says Alex Pretty ICU Nurse - Murdered by ICE, shows his photo with red highlights. Another is similar but reads Renee Nicole Good American Mom - Murdered by ICE

ICE may have lost the battle for public opinion, and its operation may be hampered by local pushback, but it has succeeded in one respect: It has instilled terror in Minnesota’s immigrant communities, forcing them into the shadows.

Read more from @gabydvj.bsky.social: buff.ly/eSAhblK

we need to talk about that Ring Super Bowl ad

India AI Impact Summit 2026 The India AI Impact Summit 2026 is a flagship global gathering hosted by the Government of India under the IndiaAI Mission, in collaboration with the Ministry of Electronics & Information TechnologyMeitY). Scheduled for 19–20 February 2026 in New Delhi, the Summit will bring together policymakers, researchers, industry leaders, startups, civil society, and international organizationsns to shape the future of inclusive and impactful AI

At the #AIImpactSummit, we (@cdt.org) will be hosting two sessions on February 16th and 18th digging into a few of these questions and more under the frame of building multilingual AI tools and how to put language speakers and communities at the fore. Please join impact.indiaai.gov.in/agenda

Social Media Companies Now Work for Governments—Not Users Content from rap videos to protest photos is being removed in the name of “national security.”

Local startups will also have to consider how to deal with overzealous gov requests while building tools that truly serves Indians and their rights, particularly when they're closer to the horse's mouth and govs have more tools to control access to info than before: foreignpolicy.com/2025/05/14/c...

Inconsistent enforcement, suppressed speech, unchecked hate mar Tamil content moderation Study reveals challenges Tamil speakers face with social media content moderation, highlighting inequities and lack of transparency.

Language capabilities matter at all levels - from ensuring AI tools work as intended to ensuring safety mechanisms don’t disproportionately burden language speakers as @thakurdhanaraj.bsky.social and I elaborated upon in the case of Tamil for The Hindu www.thehindu.com/opinion/op-e...

Government Agencies Should be Careful in Allowing Free Access to Western AI Tech Companies What is needed is meaningful consultation with the public and local language expertise available in India.

I ask similar questions in my @thewire.in piece from late 2024: Is the GoI simply giving a blank cheque of access to Big Tech under the guise of "linguistic diversity" or do language stewards and local startups have a seat at the table? thewire.in/tech/big-tec...

Can India be a “third way” AI alternative to the U.S. and China? The AI Impact Summit in India, the first in a developing country, proposes an option that focuses on public good and development. But can it deliver without capitulating to Big Tech?

Will India truly forge a 'third way' at the #AIImpactSummit next week in Delhi or further entrench Big Tech actors' roles in the Global South at the expense of local alternatives? Great read by @ambakak.bsky.social and Astha Kapoor: restofworld.org/2026/india-a...

Better Global Tools, Better Global Communities: Recommendations for Trust & Safety Teams When Building and Procuring Multilingual Tools Research has shown that automated Trust & Safety (T&S) tools often fall short in languages other than English. As part of research conducted by CDT to understand the challenges of moderating low resou...

More from me on how AI developers, particularly those working in Trust & Safety, should consider linguistic diversity and context when developing multilingual tools: cdt.org/insights/bet...

Bad Bunny, Spanish Dialects, and What to Do About the ChatGPTfication of Language Your LLM is likely missing the actual soul behind the words you speak, read, and write.

Incredible piece about what we lose when AI systems aren't trained on natively-created representative datasets: Bad Bunny lyrics!

lausworld.substack.com/p/the-limita...

Attacks on academic freedom and cutting of stable, accessible channels to immigrate to the US will have ripple effects for decades to come. The US knows little of the brain drain that's about to come and that other countries which too have faced shifts towards authoritarian power have experienced

Google Handed ICE Student Journalist’s Bank and Credit Card Numbers Amandla Thomas-Johnson didn't know how much information ICE requested in a subpoena now. Google never gave him a chance to fight it.

Alarming new revelations about ICE's use of administrative subpoenas to get information about people the administration wanted to detain and deport because of their protests against government policies. theintercept.com/2026/02/10/g...