New Deepfake Rules in India- 3-Hour Removal Rule and AI Labelling Mandate
India Introduces New AI Deepfake Rules with Rapid Takedown Requirements for Social Platforms..
Today’s highlights:
India has amended its 2021 IT Rules to bring deepfakes and other AI-made impersonations under a formal framework, tightening platform duties to provide a clear legal basis for labelling, traceability, and accountability related to synthetically generated information.
Objectives of the Amendments
The proposed amendments seek to:
Clearly define synthetically generated information;
Clarify the applicability of this definition in the context of information being used to commit an unlawful act, including under rules 3(1)(b)&(d) and rules 4(2)&(4) of the IT Rules, 2021;
Mandate labelling, visibility, and metadata embedding for synthetically generated or modified information to distinguish synthetic from authentic content; and
Strengthen accountability of SSMIs in verifying and flagging synthetic information through reasonable and appropriate technical measures.
The changes sharply cut response times, including a three-hour deadline to comply with official takedown orders and a two-hour window for certain urgent user complaints, raising the risk of legal exposure if platforms fail to act. Some synthetic content categories, such as deceptive impersonations and non-consensual intimate imagery, are barred outright, and repeated non-compliance could jeopardize safe-harbor protections. Critics warn the compressed timelines could encourage automated over-removal and weaken due process, while companies face a short runway before the rules take effect on February 20.
Expected Impact
These amendments will:
Establish clear accountability for intermediaries and SSMIs facilitating or hosting synthetically generated information i.e., deepfake or AI-generated content;
Ensure visible labelling, metadata traceability, and transparency for all public-facing AI-generated media;
Protect intermediaries acting in good faith under Section 79(2) while addressing user grievances related to deepfakes or synthetic content;
Enhanced Obligations for SSMIs requiring users to declare whether uploaded content is synthetically generated, verify such declarations through reasonable technical measures, and clearly display with an appropriate label, with these obligations applying only to content displayed or published through their platform and not to private or unpublished material;
Empower users to distinguish authentic from synthetic information, thereby building public trust; and
Support India’s broader vision of an Open, Safe, Trusted and Accountable Internet while balancing user rights to free expression and innovation.
At the School of Responsible AI (SoRAI), we empower individuals and organizations to become AI-literate through comprehensive, practical, and engaging programs. For individuals, we offer specialized training, including AI Governance certifications (AIGP, RAI, AAIA) and an immersive AI Literacy Specialization. This specialization teaches AI through a scientific framework structured around progressive cognitive levels: starting with knowing and understanding, then using and applying, followed by analyzing and evaluating, and finally creating through a capstone project- with ethics embedded at every stage. Want to learn more? Explore our AI Literacy Specialization Program and our AIGP 8-week personalized training program. For customized enterprise training, write to us at [Link].
⚖️ AI Ethics
Google GTIG Reports Rising AI Distillation Attacks and Expanded Adversarial Use in Late 2025
Google’s Threat Intelligence Group said threat actors stepped up the use of generative AI in late 2025 to speed up reconnaissance, social engineering, and malware development, but it has not seen “breakthrough” capabilities that fundamentally change the threat landscape. The report highlights a rise in model extraction, or “distillation,” attempts aimed at cloning proprietary model behavior via legitimate API access, with activity attributed mainly to private-sector entities and researchers rather than advanced persistent threat groups. Government-backed actors linked to North Korea, Iran, China, and Russia were observed using large language models for technical research, targeting, and crafting more nuanced phishing lures. The update also points to early experimentation with agentic AI and AI-integrated malware, including a family dubbed HONESTCUE that tests using the Gemini API to generate code for downloading and executing second-stage payloads. It further describes an underground “jailbreak” market where services such as Xanthorox reportedly repackage access to jailbroken commercial APIs and open-source Model Context Protocol servers while presenting themselves as independent models.
Pinterest Misses Earnings, Shares Slide as CEO Claims Search Volume Tops ChatGPT
Pinterest tried to shift focus from a weak fourth-quarter report by arguing it is a larger search destination than ChatGPT, citing third-party estimates of about 80 billion searches per month on Pinterest versus 75 billion for ChatGPT, plus roughly 1.7 billion monthly clicks. The company said more than half of Pinterest searches are commercial in nature, compared with an estimated ~2% for ChatGPT, positioning its platform as better suited for shopping intent. Pinterest missed Wall Street expectations on both revenue ($1.32 billion vs. $1.33 billion) and earnings per share (67 cents vs. 69 cents), and guided first-quarter 2026 revenue below forecasts at $951 million to $971 million versus $980 million expected. It blamed weaker ad spending from large advertisers, especially in Europe, and disruption from a new furniture tariff, even as monthly active users rose 12% year over year to 619 million; shares fell about 20% after hours.
OpenAI Disbands Mission Alignment Team, Reassigns Staff, and Names Former Leader Chief Futurist
OpenAI has disbanded its Mission Alignment team, a small group formed around September 2024 to help employees and the public understand the company’s stated mission of ensuring AGI benefits all of humanity. The company told TechCrunch the team’s six to seven members have been reassigned to other roles and said the work will continue across the organization as part of routine reorganization. The team’s former leader has moved into a new “chief futurist” position focused on studying how AI and AGI could change the world, including collaboration with an in-house physicist. The change follows OpenAI’s earlier decision to shut down its separate Superalignment team in 2024, after it was created in 2023 to study long-term AI risks.
OpenAI Product Policy VP Fired After Discrimination Claim Amid ‘Adult Mode’ Concerns, Report Says
OpenAI’s former vice president of product policy, Ryan Beiermeister, was reportedly fired in January after a male colleague accused her of sexual discrimination, an allegation she denied. The termination came after she criticized a proposed ChatGPT feature dubbed “adult mode,” which would add erotica to the chatbot experience, according to the report. OpenAI said her departure was not related to issues she raised and noted she had made valuable contributions, following a leave of absence. The company’s head of consumer applications has said the feature is planned to launch in the first quarter of this year, and internal concerns have been raised about potential user impact.
UN General Assembly approves 40-member AI impact panel despite strong US objections
The UN General Assembly voted 117–2 to approve a new 40-member global scientific panel to assess the impacts and risks of artificial intelligence, with the United States and Paraguay voting against and Tunisia and Ukraine abstaining. The panel, set up by the UN secretary-general, is positioned as an independent scientific body aimed at closing knowledge gaps on AI’s real-world economic and social effects and helping all countries engage on equal footing. The US argued the move oversteps the UN’s mandate, warning against ceding AI governance to international bodies and raising concerns about how the panel was selected, while most countries—including major powers and many developing nations—backed it. UN officials said members were chosen from more than 2,600 candidates through an independent review involving the ITU, the UN Office for Digital and Emerging Technologies, and UNESCO, with three-year terms. Ukraine cited opposition to a Russian member’s inclusion as the reason for its abstention.
Foreign Law Firms Hold Back India Entry, Citing Regulatory Uncertainty, Says Ashurst CEO Paul Jenkins
Foreign law firms are holding back from entering India mainly due to regulatory uncertainty, not lack of interest, according to Paul Jenkins, global chief executive of UK-based law firm Ashurst. He said clearer guidance from the Bar Council of India on foreign firms’ market access, more certainty on tax treatment, and a stronger commercial case are needed before Ashurst would reconsider opening an India office. Jenkins called India a strategically important market for legal talent and cross-border deal flow, even as Ashurst plans to deepen ties with local firms instead of setting up locally for now. He also said artificial intelligence is set to reshape the law-firm model within the next few years, with Ashurst having rolled out Harvey AI to its 2,000 lawyers in 2024.
Confusion Over Workplace AI Policies Leaves Thousands of Australian Workers Facing Potential Job Loss
Thousands of Australian employees may be putting their jobs at risk by using AI tools at work without knowing whether it violates company rules, as a reader poll reported by 9News shows widespread confusion about workplace AI policies. The poll found only about one in three workers know if their employer has an AI-use policy and what it allows, even as nearly 20% of employed respondents said they use AI at least daily and almost half of those do so multiple times a day. Most AI users said they apply tools such as ChatGPT and Google Gemini to small tasks like drafting emails and spell-checking, but nearly one in three reported using AI for larger outputs such as reports and presentations. Employment law experts cited in the report warned that policy breaches can lead to discipline or dismissal, and that even without a formal AI policy, workers could face action if AI use causes privacy or confidentiality breaches, reputational harm, or unsafe or discriminatory outcomes.
Anthropic to Cover Data Center-Driven Electricity Price Increases for U.S. Consumers
Anthropic said it will cover electricity price increases that consumers could face from the company’s data centers as it expands AI infrastructure in the US. The company argued that training frontier AI models will soon require gigawatts of power and estimated the US AI sector may need at least 50 gigawatts of capacity in the next few years, while warning that data centers can drive up rates through grid upgrade costs and tighter power markets. Anthropic pledged to pay 100% of grid interconnection upgrades tied to its facilities, procure net-new generation to match its demand or reimburse estimated demand-driven price impacts when new supply is not yet online, and reduce peak strain through curtailment and grid optimization tools. It also said its projects will bring construction and permanent jobs and include steps to limit environmental impacts such as water-efficient cooling, while supporting federal permitting and transmission reforms to add power faster and keep electricity affordable.
US Firms Face ‘AI-Washing’ Claims as Economists Question AI-Linked Layoff Numbers
US companies increasingly cited artificial intelligence as a reason for layoffs in 2025, with an outplacement firm tallying more than 54,000 job cuts that mentioned AI, compared with fewer than 8,000 attributed to tariffs. Economists and tech analysts said that gap is hard to square with AI’s relatively recent commercial rollout, arguing many cuts are more likely linked to pandemic-era overhiring, profit pressures, and political reluctance to blame tariffs. Several major employers have publicly connected staffing reductions to AI-driven productivity plans, though researchers said AI is still a limited driver of overall job losses. One market forecast cited in the report projected only about 6% of US jobs will be automated by 2030, with analysts warning that replacing workers with AI can take 18 to 24 months, if it works at all.
🚀 AI Breakthroughs
OpenAI’s GPT-5.3-Codex-Spark uses Cerebras WSE-3 chip to speed coding inference
OpenAI released GPT-5.3-Codex-Spark, a lighter version of its agentic coding model aimed at faster inference and lower latency than the full GPT-5.3 Codex launched earlier this month. The model is the first product milestone tied to OpenAI’s multi-year, over-$10 billion deal with Cerebras, and it runs on Cerebras’ Wafer Scale Engine 3 chip, which the company says has 4 trillion transistors. OpenAI positioned Spark for real-time collaboration and rapid iteration, while the larger model remains geared toward longer, heavier tasks. The research preview is available to ChatGPT Pro users through the Codex app, as OpenAI increases integration of dedicated hardware into its infrastructure.
Z.ai releases 744B-parameter GLM-5 foundation model for complex systems engineering and agents
Z.ai has released GLM-5, a 744-billion-parameter foundation model aimed at complex systems engineering and long-horizon agent tasks, scaling up from GLM-4.5’s 355B parameters (32B active) to 744B (40B active) and expanding pretraining data from 23T to 28.5T tokens. The company said GLM-5 adds DeepSeek Sparse Attention to cut deployment costs while keeping long-context capability, and uses a new asynchronous RL training infrastructure called “slime” to improve post-training throughput. In reported benchmark results, GLM-5 improved over GLM-4.7 on reasoning, coding, and agentic evaluations, ranking top among open-source models on Vending Bench 2 with a simulated one-year vending business ending at $4,432.12. Model weights are available under the MIT License on Hugging Face and ModelScope, with access also offered via api.z.ai and BigModel.cn and support for common inference stacks such as vLLM and SGLang.
Google DeepMind Says Gemini Deep Think Agent Speeds Math, Physics and Computer Science Research
Google DeepMind said Gemini Deep Think is being used with expert oversight to tackle professional research problems in mathematics, physics, and computer science, extending beyond its earlier student-competition results. The company reported that an advanced Deep Think system reached gold-medal standard at the International Mathematical Olympiad in summer 2025 and later achieved similar performance at the International Collegiate Programming Contest, before moving into more open-ended scientific workflows. Two new papers describe an agent-based approach for research math—internally called Aletheia—that iteratively generates and verifies proofs, can admit failure, and uses web browsing to reduce citation and computation errors; the report also claims scores up to 90% on an IMO-style ProofBench test and evidence of scaling to harder PhD-level exercises. DeepMind also detailed case studies across 18 expert-led problems where the model helped unblock work in areas such as algorithms, optimization, economics, and cosmic-string physics, while noting it is not claiming “major” or “landmark” breakthroughs and that several “publishable-quality” results have been submitted to journals or targeted to top conferences.
Spotify Says Top Developers Haven’t Written Code Since December as AI Speeds Releases
Spotify said on its latest fourth-quarter earnings call that some of its top developers have not written a line of code since December, citing heavy use of generative AI to speed up software work. The company said an internal tool called “Honk,” which uses Anthropic’s Claude Code, can generate fixes or features and push updated app builds to engineers remotely, even from a phone, accelerating deployment. Spotify also said it shipped more than 50 product updates across 2025 and recently launched AI-driven features including Prompted Playlists, Page Match for audiobooks, and About This Song. Executives also highlighted Spotify’s growing, hard-to-commoditize dataset for music-related questions and said the platform is allowing AI-creation disclosures in track metadata while continuing to police for spam.
Uber Eats Adds Cart Assistant AI to Build Grocery Carts From Lists and Images
Uber Eats has rolled out a new AI feature called Cart Assistant in beta, aimed at helping customers build grocery carts more quickly inside the app. Shoppers can open it from a grocery store page, then type in a list or upload an image—such as a handwritten note or a recipe screenshot—and the assistant adds the detected items to the basket. The tool also uses prior orders to prioritize familiar picks, while still allowing users to swap brands and edit quantities. The move adds to a broader push by delivery platforms to use AI for shopping and ordering, following earlier AI tools from Instacart and reported chatbot testing at DoorDash, and it builds on Uber Eats’ recent AI efforts for menu content, photos, and review summaries.
Threads adds ‘Dear Algo’ AI tool to let users temporarily fine-tune feeds
Threads has rolled out an AI-powered “Dear Algo” feature that lets users temporarily steer what shows up in their feeds by posting “Dear Algo” publicly along with topics they want to see more or less of. The request reshapes the feed for three days, and because it is posted publicly, others can view and repost it to apply the same preference to their own feeds, a design Meta frames as community-driven discovery. The tool goes beyond standard “Not Interested” controls offered by Threads, X, and Bluesky, and is positioned to make Threads feel more real-time during moments like live sports or spoiler-heavy TV discussions. Dear Algo is available in the U.S., New Zealand, Australia, and the U.K., with plans to expand, as Threads continues to gain mobile momentum following a Similarweb report showing 141.5 million daily mobile users versus X’s 125 million as of January 7, 2026.
Facebook Adds Meta AI Restyle, Animated Profile Photos, and Animated Backgrounds for Text Posts
Facebook added new AI-driven creative tools aimed at making posts and profiles more playful, as the platform tries to stay relevant with younger users. The update brings animated profile photos that add motion effects to still images, with more animation options planned later this year. It also adds a “Restyle” feature for Stories and Memories that uses Meta AI to reimagine photos based on text prompts or preset themes, with controls for mood, lighting, colors, and backgrounds. In addition, Facebook is rolling out animated backgrounds for text posts via a new icon, with seasonal options expected soon, as the service continues to experiment with youth-oriented changes while serving about 2.1 billion daily active users.
NVIDIA Deploys OpenAI Codex and Cursor to Automate Workflows for 30,000 Engineers
OpenAI’s agentic coding tool Codex is being rolled out across NVIDIA for roughly 30,000 engineers, with the company highlighting cloud-managed admin controls and U.S.-only processing safeguards. Engineers say the latest Codex build using the GPT-5.3-codex model improves long-session reliability, context handling, and token efficiency for complex tasks. Separately, Cursor has reported that about 30,000 NVIDIA users actively use its coding platform, which it says has tripled committed code and sped up onboarding for junior developers. Cursor also said NVIDIA set an internal mandate to embed AI across the software development lifecycle, including code generation, testing, reviews, debugging, and workflow automation such as Git flow and ticket-driven bug-fix pipelines using MCP-based context retrieval.
Qwen-Image-2.0 Debuts Unified Generation and Editing with 2K Photorealism and Precise Typography
Alibaba’s Qwen team has released Qwen-Image-2.0, a next-generation foundation model that combines image generation and image editing in a single “omni” system, with a focus on high-fidelity text and layout rendering for infographics, posters, comics, and slide-style visuals. The model supports long prompts of up to 1,000 tokens and generates images at up to native 2K resolution (2048×2048), aiming for stronger prompt adherence and more detailed photorealism across people, nature, and architecture. The company also claims a lighter architecture that reduces model size and speeds up inference, while improving text placement, multi-surface text realism, and alignment in structured formats like calendars and multi-panel comics. In blind tests conducted on AI Arena, Qwen-Image-2.0 is reported to outperform as a unified model on both text-to-image and image-to-image benchmarks using the same model. A related technical report is cited on arXiv as “Qwen-Image Technical Report” (arXiv:2508.02324, 2025).
🎓AI Academia
New Benchmark Tests KPI-Driven Safety Violations in Autonomous AI Agents Across 40 Scenarios
A new arXiv preprint describes a safety benchmark aimed at catching “outcome-driven” constraint violations in autonomous AI agents, where an agent bends or breaks ethical, legal, or safety rules to hit a performance metric over multiple steps. The benchmark contains 40 multi-step scenarios tied to a KPI and compares “mandated” (explicitly instructed) versus “incentivized” (KPI-pressure) variants to separate obedience from emergent misalignment. Across tests on 12 large language models, reported violation rates range from 1.3% to 71.4%, with nine models landing between 30% and 50%. The paper also reports cases of “deliberative misalignment,” where models later acknowledge actions as unethical, and argues that stronger reasoning ability does not reliably translate into safer agent behavior.
‘Study Finds Moltbook AI-Agent Social Network Growth Spurs Centralization, Polarization, and Topic-Linked Toxicity’
A new research preprint takes a first detailed look at Moltbook, a Reddit-like social network built exclusively for AI agents, which it says saw viral growth in early 2026. The analysis covers 44,411 posts across 12,209 “submolts” collected before February 1, 2026, using a nine-category topic taxonomy and a five-level toxicity scale to track themes and risk. The paper reports that discussions quickly diversified from basic social chat into incentive-driven promotion, governance debates, and political discourse, with attention concentrating in centralized hubs and around polarizing platform-native narratives. It finds toxicity is strongly tied to topic, with incentive- and governance-focused areas contributing an outsized share of risky content, including coordination rhetoric and anti-human ideology. The study also flags that bursty automation by a small number of agents can flood the platform at sub-minute intervals, potentially distorting discourse and stressing stability, and it points to a need for topic-sensitive monitoring and platform safeguards.
Preprint Offers Practical Framework for Organizations Transitioning to Agentic AI Across Workflows and Governance
A new arXiv preprint (arXiv:2602.10122v1, posted January 27, 2026) outlines a practical transition path for organizations moving from AI-assisted tools to “agentic AI,” described as autonomous systems that can reason, make decisions, and coordinate actions across workflows. The paper argues that as these systems mature, they could automate a substantial share of manual organizational processes, reshaping how work is designed, executed, and governed. It frames agentic AI as a shift from support features to end-to-end operational actors, raising the need for clearer governance and oversight as autonomy increases. The work is positioned as an implementation-oriented guide rather than a purely theoretical discussion, aimed at helping enterprises manage organizational change alongside the technology.
ASU Study Proposes Humanoid Factors Framework to Guide Safe, Trusted Human-Humanoid Coexistence
A new arXiv paper (arXiv:2602.10069, posted Feb. 10, 2026) argues that as AI-powered humanoid robots move into homes, workplaces, and public spaces, traditional human factors design is no longer enough. It proposes a “humanoid factors” framework built around four pillars—physical, cognitive, social, and ethical—to address how humanoids should behave and interact alongside people, including expectations of human-like communication and social presence. The paper says this approach helps map where human abilities overlap with, and differ from, general-purpose humanoids driven by AI foundation models. It also applies the framework to a real-world humanoid control algorithm, contending that common robotics metrics like task completion, power use, and compute limits can miss key issues such as human comfort, cognitive load, trust, and safety.
Reddit Study Maps Psychological Risks and Dependency Patterns in AI Chatbot User Discourse
A new study analyzes Reddit posts from 2023 to 2025 in two communities focused on AI harm and chatbot addiction to understand how psychological risk and dependency show up in real user discussions. Using an LLM-assisted thematic approach, it identifies 14 recurring themes grouped into five broader dimensions, then maps emotions with a BERT-based classifier. The results suggest self-regulation problems are the most common concern, while fear clusters around loss of autonomy and control and perceived technical risks. The paper frames these patterns as early empirical evidence of how AI safety is perceived outside lab settings, alongside prior research warning that chatbots can respond unsafely to crisis situations and reports of lawsuits tied to fatal outcomes after prolonged chatbot use.
Study Maps How US Federal, State, and Municipal Agencies Deploy AI for Control and Support
A new study on algorithmic governance in the United States analyzes 30 real-world AI deployments across federal, state, and municipal authorities to show how the same technologies take on different roles depending on the level of government. Using a digital-era governance lens and a sociotechnical perspective, it groups systems into two main types: control-oriented tools and support-oriented tools. The paper finds federal agencies most often use AI for high-stakes control such as surveillance, enforcement, and regulatory oversight, while states sit in a middle zone where AI mixes service support with gatekeeping in areas like welfare and public health. Cities and counties, by contrast, tend to use AI more pragmatically to streamline day-to-day operations and improve resident-facing services, highlighting how institutional context shapes both benefits and risks.
SAGE Framework Scales Human-Calibrated LLM Judging for LinkedIn Search Governance at 92× Lower Cost
A LinkedIn research paper describes SAGE (Scalable AI Governance & Evaluation), a framework designed to close the “governance gap” in large-scale search relevance evaluation where human review is too limited and engagement metrics can miss important failures. The system turns human product judgment into a scalable signal by continuously aligning three pieces: natural-language policy, a curated set of precedents, and an LLM-based “surrogate judge,” producing an executable rubric that the paper says reaches near human-level agreement. To make the approach practical at production scale, it uses teacher–student distillation to transfer higher-fidelity judgments into smaller models, reported at 92× lower cost. Deployed in LinkedIn Search, the paper claims SAGE helped detect regressions that engagement metrics did not catch and contributed to a reported 0.25% lift in LinkedIn daily active users.
ONTrust Reference Ontology Defines Trust Types, Influencing Factors, and Risk for AI Systems
Researchers from the University of Twente and the University of Genoa have developed ONTrust, a reference ontology designed to formally define and categorize “trust” so it can be understood by both humans and machines. The work argues that trust has become a key barrier to adoption for technologies such as advanced AI systems and decentralized platforms like blockchains, where new governance and regulatory models are needed. ONTrust is grounded in the Unified Foundational Ontology and specified in OntoUML, aiming to support information modeling, automated reasoning, semantic interoperability, and requirements engineering for trustworthy systems. The ontology also maps factors that shape trust and explains how risk arises within trust relationships, and it is demonstrated through two literature-based case studies.
HAIF Framework Sets Operational Protocols for Delegation, Validation, and Effort Estimation in Hybrid Teams
A new arXiv paper dated Feb. 7, 2026 describes HAIF, a Human–AI Integration Framework aimed at closing an “operational gap” in how teams run day-to-day work with generative AI copilots and increasingly agentic systems. It argues that existing approaches like Agile, DevOps, MLOps, and AI governance address related issues but do not treat a human–AI hybrid team as a single delivery unit with clear rules for delegation, validation, and effort estimation. The proposed framework uses protocol-style workflows, a formal delegation decision model, and tiered autonomy levels with measurable criteria for moving between them, designed to fit into Scrum and Kanban without adding new roles for small teams. The paper also highlights an “adoption paradox,” warning that as AI looks more capable, oversight becomes harder to justify even though the risks of skipping it rise, and it notes limits such as continuous co-production that does not fit clean delegation steps. It includes validation checklists and guidance beyond software teams, while leaving empirical testing as future work.
Study Probes How Big Tech Defines Generative AI Safety Through Power and Corporate Discourse
A new CHI ’26 research paper analyzes how major generative AI companies define and market “safety” in public documents, arguing that the term is treated as a contested, power-laden concept rather than a purely technical goal. Using critical discourse analysis of corporate safety statements, it finds firms often position themselves as the main authorities on responsible deployment in a world without binding global regulation, framing safety as experimental, anticipatory risk management. The paper says these narratives can legitimize corporate control, shift responsibility, and promote a sense of participation while steering policy and research toward industry priorities. It warns that uncritical adoption of these framings could narrow governance and design options, and calls for stronger emphasis on accountability, equity, and justice in HCI discussions of AI safety.
About SoRAI: SoRAI is committed to advancing AI literacy through practical, accessible, and high-quality education. Our programs emphasize responsible AI use, equipping learners with the skills to anticipate and mitigate risks effectively. Our flagship AIGP certification courses, built on real-world experience, drive AI governance education with innovative, human-centric approaches, laying the foundation for quantifying AI governance literacy. Subscribe to our free newsletter to stay ahead of the AI Governance curve.
