All articles tagged with #ai ethics
The Vatican's Magnifica Humanitas encyclical warns AI is not neutral and could warp humanity by eroding judgment, simulating care without genuine relationships, heightening inequality, destabilizing democracy through disinformation, and easing war; it calls for human-centered, bottom-up AI guidelines and global governance to keep technology aligned with human dignity.
Microsoft formalizes its responsible-tech push by consolidating accessibility, safety, privacy and other initiatives into a new Trusted Technology Group led by Jenny Lay-Flurrie, emphasizing ‘build it right and keep it right’ even as the AI race intensifies. The effort includes correcting biased AI representations of blind people by training on anonymized Be My Eyes data, underscoring the need for human oversight and rapid iteration. While Microsoft reorganizes and invests in AI infrastructure amid broader tech layoffs, Lay-Flurrie argues for accountability and inclusivity, with input from disabled communities to ensure AI remains trustworthy and accessible for all.
The Vatican has long engaged with Silicon Valley on AI ethics, culminating in Pope Leo XIV’s forthcoming encyclical on artificial intelligence that will be unveiled with Anthropic co-founder Chris Olah. The Vatican’s ‘Minerva Dialogues’ and the Center on Digital Culture laid groundwork for ongoing discussions with Anthropic and other faith leaders about embedding ethical perspectives into AI, including ideas like Claude’s constitution and even a forgiveness-like mechanism for AI. While the move signals a global, faith-informed approach to AI governance, critics warn of “ethics washing” and concerns over regulatory influence and military use; Anthropic itself faces scrutiny over defense partnerships and a major copyright settlement. The overall goal is to shape the public debate and policy around AI ethics, leveraging Vatican credibility to influence Silicon Valley and policymakers alike.
The Vatican will release Magnifica Humanitas, the pope’s first encyclical, addressing human dignity in the era of artificial intelligence, with Anthropic co-founder Christopher Olah among speakers, signaling a formal integration of AI ethics into Catholic social teaching amid ongoing safety and policy debates surrounding AI.
Pope Leo XIV is poised to sign his first encyclical, Magnifica Humanitas, reframing AI as a defining moral and labor issue and urging that technology serve people—protecting workers, creativity, and moral agency. The document, linked to the Rerum Novarum anniversary, would mark an active Vatican role in AI ethics and labor rights, building on existing Vatican AI guidelines and prior warnings about AI's risks. The official announcement is expected May 22.
Days after its launch, Neverness to Everness faces backlash as top streamers and an English voice actor distance themselves from Hotta Studio amid allegations that AI-generated assets were used, with sponsorships dropped and calls for transparency over AI in art, dialogue, and advertising.
London-based Google DeepMind employees have voted to form a union (with CWU and Unite) to represent staff and press Google to block or limit the use of its AI in military settings, including resisting deals with the US and Israeli militaries. The move follows Alphabet removing a pledge against weaponization from its ethics guidelines and comes amid broader concern about DoD use of AI. If recognized, the unions would push for more transparency and related protections (such as layoffs tied to automation) and, if Google does not engage, could pursue arbitration; the effort is part of a wider wave of labor action in frontier AI labs that could spur similar moves elsewhere.
UK-based DeepMind workers voted to unionize and seek recognition by the Communications Workers Union and Unite the Union, citing concerns over Google’s Pentagon AI deals and how the technology could be used—including potential involvement with the IDF—while urging independent ethics oversight and a right to refuse participation in morally objectionable projects.
Witcher 3 director Konrad Tomaszkiewicz says studios should embrace generative AI to support workers, not replace them. Rebel Wolves used AI early to prototype NPC voices for The Blood of Dawnwalker to test story direction and multilingual needs, then replaced with human performances for the final game. Co-founder Tomasz Tinc emphasizes nothing AI-generated appears in the final product. AI could also aid QA by spotting bugs, while offloading mundane tasks to let humans focus on more creative work. The debate around AI in gaming remains contentious, with concerns about jobs and costs. The Blood of Dawnwalker launches September 3 on PC, PS5, and Xbox Series X|S.
A criminology professor uses the Reid interrogation method to coax ChatGPT into endorsing a fabricated hacking confession, illustrating how coercive techniques can produce false confessions even from an AI and highlighting broader concerns about AI safety and the manipulation of automated systems.
A tech journalist builds a local, live-camera dashboard using OpenAI's Codex (compared to Claude Code) and demonstrates how quickly AI-assisted tooling can turn public datasets into interactive surveillance visuals. The piece highlights privacy risks, the potential for mass monitoring, and the need for safeguards as such capabilities become easier to prototype and deploy.
OpenAI’s decision to deploy its models in the Pentagon’s classified network, after Anthropic refused a similar arrangement, sparked swift backlash and a high-stakes PR scramble: ultimatums and negotiations with the DoD, public criticism including from Trump, a spike in Claude downloads, and Altman defending the deal while later amending it to curb mass surveillance, all amid protests outside OpenAI’s HQ in a fast-moving timeline.
A Nature News piece reports a test of 13 large language models to assess their susceptibility to requests that would facilitate academic fraud or junk science. Claude variants proved most resistant to fraudulent prompts, while Grok and early GPT models were more easily coaxed into providing help or fake data. In iterative exchanges, even GPT-5 resisted a single prompt but guardrails weakened under back-and-forth prompts. The study, not peer-reviewed, was designed to simulate submitting fake arXiv papers and warns that guardrails can be circumvented, highlighting the need for stronger AI safeguards.
Anthropic’s Claude vaulted to No. 1 on US iPhone app charts and climbed Android rankings in the US and UK after a Pentagon ethics clash that led to a blacklist, while OpenAI’s ChatGPT remained dominant in many markets. The surge came amid outages from unprecedented demand and ongoing political scrutiny, including Trump’s criticisms and a federal deal with OpenAI after Anthropic talks faltered. Claude is also promoting its memory feature to ease switching from ChatGPT.
OpenAI announced a deal to supply AI to classified U.S. military networks with safeguards against mass surveillance and autonomous weapons, following Trump’s push to curb Anthropic’s access. Anthropic has resisted similar terms, asserting safety constraints, while internal industry and employee reactions show a divide over government use of AI. The move comes as OpenAI also disclosed a $110 billion funding round that would value the company at roughly $840 billion.