GLAAD Says AI Is Failing LGBTQ Customers—And Warns the Danger Is Rising – Decrypt

Synthetic intelligence is amplifying anti-LGBTQ bias, misinformation, and discrimination in methods that may have an effect on all the things from healthcare and employment to housing and privateness, in accordance with a brand new report from advocacy group GLAAD.

Launched on Wednesday, the report titled “Construct for Everybody: A Framework for LGBTQ Illustration and Security in AI,” argues that LGBTQ security must be handled as a core requirement of accountable AI improvement.

GLAAD warns that AI skilled on biased or incomplete information can reinforce stereotypes, suppress LGBTQ voices, expose customers to privateness dangers, and produce discriminatory outcomes because the know-how turns into more and more embedded in on a regular basis life.

“AI is a civil rights difficulty,” GLAAD President and CEO Sarah Kate Ellis wrote within the report. “Neutrality is not an choice. To construct AI that’s moral, inclusive, and accountable, tech leaders should proactively embrace intentional practices to create protected merchandise.”

Areas of concern, the examine stated, embrace biased coaching information, anti-LGBTQ misinformation, discriminatory outcomes in predictive AI methods, content material moderation failures, and privateness dangers, arguing that AI methods skilled on incomplete or inaccurate details about LGBTQ individuals can reinforce stereotypes.

Ellis argued that accountable AI is finest for enterprise and a requirement for future-proofing AI firms. “Greater than 20 p.c of Gen Z is LGBTQ,” she wrote. “These are your future staff and shoppers.”

In accordance with a 2023 examine by advisory and funding agency LGBT Capital, the worldwide shopping for energy of LGBTQ individuals is $4.7 trillion, with that quantity estimated to achieve $33 trillion by 2030.

“To place that in perspective, if we had been a rustic, we’d be the 4th largest economic system on the earth,” Ellis wrote.

The report comes amid an ongoing debate over AI bias. In Could, researchers discovered main AI fashions persistently favored Catholicism whereas responding much less favorably to Jehovah’s Witnesses, atheism, and agnosticism.

Earlier this month, former xAI engineer Devin Kim sued xAI and SpaceX, alleging he was fired after warning that Grok lacked satisfactory safeguards in opposition to misinformation and bias. In the meantime, the Elon Musk-led xAI is preventing a authorized battle in opposition to Colorado over a state legislation requiring firms to evaluate and scale back discrimination dangers in AI methods used for selections involving housing, employment, and lending.

GLAAD argues the results prolong past chatbot conversations and picture mills.

“Whereas not particular to LGBTQ individuals, we should additionally point out different rising challenges as AI improvement and adoption progresses,” the examine stated. “These can embrace mannequin hallucinations or sycophantic conduct that generate misinformation, together with about consequential matters corresponding to well being or elections.”

These issues turn into extra important, the examine stated, as firms push AI brokers able to performing duties with restricted human oversight. GLAAD warned that autonomous brokers might inherit present biases and automate discriminatory outcomes, corresponding to excluding LGBTQ-affirming healthcare suppliers from search outcomes or making incorrect assumptions about customers’ identities.

To forestall these dangers from changing into additional embedded in AI methods, GLAAD referred to as on builders to enhance LGBTQ illustration in coaching information, strengthen privateness protections, preserve human oversight of moderation methods, and work extra intently with advocacy teams. The report additionally requires stronger business accountability and regulatory oversight.

“Failure to account for LGBTQ experiences and points in coaching information, product design, and governance may result not solely in hurt to marginalized communities but in addition in inaccurate, lower-quality merchandise which will undermine consumer belief in a rising demographic,” the examine stated.