May 28 Roundup: Illinois Passes America's Toughest AI Safety Law, OpenAI Fights Election Misinformation, Meta Goes Paid, and DuckDuckGo Surges 30% Against Google's AI Search
States are stepping up where Congress won't. A watershed AI safety law just cleared Illinois. OpenAI is waging quiet cyberwar ahead of elections. Meta wants your money for its AI. And millions of users are fleeing Google's AI-first search — straight to DuckDuckGo. Here's everything that moved the needle in AI yesterday.
1. Illinois Passes America's Strongest AI Safety Law — Requiring Independent Audits of OpenAI, Anthropic, and Google
In a move that immediately rewrote the playbook for AI accountability, the Illinois House of Representatives passed SB 315 on Wednesday — a bill that goes further than any existing state law by requiring frontier AI companies like OpenAI, Anthropic, and Google DeepMind to have their safety practices verified by independent third-party auditors. Governor JB Pritzker announced via social media that he intends to sign the bill.
What sets Illinois apart from California and New York — which require AI companies to publish reports and disclose guardrails — is the independent verification component. Under SB 315, companies can no longer simply grade their own safety homework. An outside auditor must confirm that an AI lab is actually adhering to its own published safety standards. Previously, no independent body was required to hold labs to their own claims.
"We're in a situation where the AI companies grade their own homework," one safety advocate told WIRED. "This changes that."
The timing is significant. Congress has passed essentially no meaningful AI safety legislation despite years of hearings, leaving state legislatures as the primary arena where AI accountability is actually being forged. Illinois' bill arrives as voters across the political spectrum consistently tell pollsters they want more oversight of AI companies — not less. OpenAI's chief of global affairs, Chris Lehane, told WIRED last week that the company's AI policy is now explicitly oriented around passing a series of similar state laws, suggesting OpenAI may prefer a patchwork of state-level rules over a single federal framework it can't negotiate.
The backdrop also includes a notable federal retreat: President Trump last week reportedly scuttled a planned executive order on AI regulation at the last minute, citing concerns it might "get in the way" of U.S. competitiveness against China. Governor Newsom, meanwhile, has been moving in the opposite direction, deepening California's AI controls and signaling a possible 2028 presidential run with AI accountability as a central theme.
Independent safety audits are a genuine step change. Until now, "responsible AI" has largely meant self-attestation — companies publishing safety cards and principles with no outside check. Illinois just made that insufficient. For enterprises buying AI services, this creates a new due-diligence expectation: if Illinois is requiring it of AI vendors, boards and legal teams will start asking why your vendor contracts don't include the same. Expect this model to spread to other states quickly, and expect AI vendors to begin proactively commissioning audits as competitive differentiation — and proof that the self-policing era is ending.
2. OpenAI Quietly Builds Election Cyber Defense — Partnering With State Officials Before the 2026 Midterms
OpenAI is announcing a package of new partnerships and initiatives aimed at combating AI-driven election misinformation ahead of the 2026 U.S. midterm elections and international votes. According to Axios, the company is offering its cybersecurity products to state election officials, backing legislation targeting AI-generated disinformation, and formalizing partnerships with election integrity organizations.
The initiative is notable for several reasons. It represents OpenAI's most explicit engagement with the electoral process — a sensitive area the company has historically approached cautiously, given that its own models are capable of generating convincing synthetic political content at scale. By proactively offering cybersecurity tools to state election officials and backing anti-disinformation legislation, OpenAI is essentially acknowledging the dual-use risk of its own technology while trying to be on the right side of the issue before a major election cycle.
The announcement lands alongside a separate Axios report revealing that OpenAI and Anthropic CEOs have been publicly diverging on the AI jobs question — with Sam Altman sounding increasingly optimistic about AI's economic impact even as Anthropic's Chris Olah, speaking at the Vatican's AI ethics conference alongside Pope Leo XIV, doubled down on warnings about existential risk and labor displacement.
"OpenAI is announcing new partnerships to combat misinformation, offering its cybersecurity products to state officials and backing legislation ahead of elections in the U.S. and globally." — Axios
OpenAI's election initiative reads as both genuine civic responsibility and smart reputation management. The company's models are already being used — and misused — in political communication. Getting ahead of that with state partnerships and legislative backing helps frame OpenAI as a solution rather than a threat. But the harder question is enforcement: offering tools is different from deploying them effectively at the state level, where capacity and expertise vary dramatically. Watch for whether these partnerships translate into actual detection and takedown capability before November, or whether this is mostly positioning. Either way, AI vendors entering any government or media sector should expect election-related use-case scrutiny to intensify over the next six months.
3. Meta Launches Global Paid AI Subscriptions — And Rolls Out Private "Incognito" AI Chat
Meta is officially in the subscription business. The company announced on Wednesday the global rollout of paid subscription tiers for Instagram, Facebook, and WhatsApp — and has begun testing additional subscription plans for businesses, creators, and Meta AI power users. The move marks a significant monetization pivot for a company that has historically relied almost entirely on advertising revenue.
Alongside the subscription launch, Meta quietly expanded its "Incognito Chat" feature for WhatsApp — a private processing mode for conversations with Meta AI that limits data storage and retention. According to multiple reports, conversations in incognito mode are designed to disappear after the chat ends and are not stored on Meta's servers. WhatsApp users can activate incognito sessions by tapping a new icon in one-on-one chats with Meta AI. The feature extends WhatsApp's end-to-end encryption ethos into AI interactions, positioning Meta as more privacy-respecting than cloud-first AI assistants that retain conversation history.
TechCrunch reports that the subscription rollout is global and covers Meta's flagship apps with more tiers coming — including AI-specific plans. For businesses, Meta is testing subscription models that would provide enhanced access to AI-powered marketing and commerce tools across its platforms.
"Meta is doubling down on its subscription offerings. On Wednesday, the social networking giant announced it's now rolling out its consumer subscription plans globally for its flagship apps, Instagram, Facebook, and WhatsApp, and beginning tests of new subscriptions for businesses, creators, and Meta AI users." — TechCrunch
Meta's subscription move is the most significant business model shift in social media since the ad-supported model became dominant. For businesses that rely on Meta's platforms for customer reach — which is nearly every consumer brand — this creates new pressure. Premium AI features behind a paywall means your competitors who pay get better tools for targeting and content creation. And the incognito AI chat feature is a genuine differentiator: users who are concerned about AI companies retaining their conversations now have a mainstream option that doesn't require switching to a privacy-first niche tool. Businesses building on WhatsApp for Business should pay close attention to how incognito mode interacts with message analytics — you may start losing visibility into AI-assisted conversations.
4. DuckDuckGo Surges 30% as Users Flee Google's AI Search Overhaul
Google's decision to replace traditional blue-link search results with AI-generated summaries and agents at its I/O conference earlier this month has triggered a measurable consumer backlash — and DuckDuckGo is the primary beneficiary. App installs for the privacy-focused search engine surged an average of 20.8% week-over-week in the seven days after Google's May 19 announcements, peaking at 37.6% growth on May 26. On iOS in the U.S., installs climbed an average of 33% during the same period, nearly touching 70% growth at peak. Visits to DuckDuckGo's AI-free search page (noai.duckduckgo.com) rose 22.7% on average week-over-week.
The numbers tell a clear story: a substantial segment of users — particularly those who rely on Google for research, comparison shopping, and professional queries — feel that AI-generated summaries are degrading the quality and transparency of search results. The shift from "here are sources" to "here is an answer (trust us)" is jarring for users who have built workflows around traditional search hygiene.
Google is aware of the tension. The company simultaneously introduced "Preferred Sources" into AI Overviews and AI Mode, allowing websites publishing fresh content to be featured in AI-powered search results. This is a direct attempt to reassure publishers and businesses that AI search doesn't eliminate organic visibility — it just changes how it works. Whether that's enough to stem the DuckDuckGo tide remains to be seen.
"Google overhauled Search at I/O 2026, replacing blue links with AI agents. The backlash has been swift. DuckDuckGo app installs spiked 30% as users seek a way out." — TechCrunch
For businesses, this is the most important search story of the year. Google's AI Mode is not optional — it's becoming the default experience. That means your entire SEO strategy, content library, and digital marketing funnel was built for an interface that is now being replaced. The "Preferred Sources" feature is Google's olive branch to publishers, but the selection criteria are opaque and the traffic dynamics are completely different from traditional ranking. Businesses need to start testing their visibility in AI Mode right now — not in Q4 planning. Separately, the DuckDuckGo spike signals a latent user demand for "no-AI" search options. If you have audiences that skew toward privacy-conscious or technically sophisticated users, consider whether your content is optimized for traditional search formats these users still prefer.
5. AI CEO Jobs Pivot: Altman and Amodei Walk Back the Apocalypse — MIT Tech Review Adds Nuance
One of the more quietly consequential stories developing this week: both Sam Altman (OpenAI) and Dario Amodei (Anthropic) have been publicly walking back their own previous predictions about AI-driven mass job displacement. AI Magazine notes that both CEOs spent the past two years actively warning that generative AI would dismantle white-collar employment — warnings that shaped public discourse, legislative hearings, and corporate workforce planning. In May 2026, the narrative has abruptly shifted toward a more optimistic framing.
MIT Technology Review's analysis adds important texture. The publication found that employment decreases are specifically concentrated in jobs "where tasks could be automated with minimal human involvement" — roles like software developers and knowledge workers performing structured, repeatable tasks. The displacement is real and measurable, but it's more targeted than the "all white-collar jobs at risk" framing that dominated 2024 and 2025 headlines.
Meanwhile, Anthropic co-founder Chris Olah, speaking at the Vatican's AI ethics conference alongside Pope Leo XIV (who issued a landmark encyclical titled "Magnifica Humanitas," or "Magnificent Humanity," on AI's role in society), doubled down on the existential-risk framing even as Altman has been talking up abundance. The divergence is striking: the two companies most associated with frontier AI safety are now publicly at odds on whether AI represents a near-term jobs catastrophe.
"The fear of AI and automation displacing human workers has been a defining anxiety for the past couple of years. This panic was actively fuelled by OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei, who repeatedly warned that Gen AI would dismantle white-collar employment. Cue May 2026 and the narrative has abruptly shifted." — AI Magazine
The CEO narrative shift matters enormously for enterprise planning. When the founders of OpenAI and Anthropic were loudly predicting job displacement, boards used that as cover to accelerate AI adoption and defer headcount decisions. Now that the same founders are pulling back, executives need to be careful not to overcorrect in either direction. MIT's data is the most useful signal here: displacement is real but concentrated in specific task categories. The strategic move is to get granular — map which specific job functions in your organization have high task-automation overlap, and build a transition plan that doesn't require waiting for the CEOs to settle on a narrative. The window for proactive reskilling investment is now, not after the next forecast revision.
6. Anthropic Expands Project Glasswing — Claude Security Goes Public Beta for Cybersecurity Teams
Anthropic expanded Project Glasswing this week with the public beta launch of Claude Security, a set of purpose-built AI tools for cybersecurity professionals. The offering includes capabilities for scanning codebases, triaging vulnerabilities, and generating remediation code — all within a framework designed for security team workflows. Anthropic is also sharing research and providing support for open-source defenders as part of the initiative.
Project Glasswing has been Anthropic's most visible bet on AI as a defensive cybersecurity tool — a direct response to growing concerns (some of which Anthropic itself has articulated) that frontier AI models like Mythos could be weaponized to discover and exploit vulnerabilities at scale. By getting Claude Security into the hands of security teams in public beta, Anthropic is simultaneously advancing the defensive use case and collecting real-world data on how AI-assisted security tooling performs in production environments.
The launch arrives in a climate where AI-assisted vulnerability exploitation is no longer theoretical. Multiple cybersecurity firms and government agencies have flagged evidence of AI-accelerated attack toolchains in the wild, making the defensive AI tooling race increasingly urgent for enterprises that can't rely on patching speed alone.
"Anthropic expands Project Glasswing with Claude Security in public beta and new cyber verification tools for eligible security teams, helping scan codebases, triage vulnerabilities, and generate fixes while sharing research and support for open-source defenders." — Releasebot / Anthropic
AI-powered vulnerability scanning is becoming a table-stakes security capability, not a premium add-on. Claude Security in public beta gives security teams a concrete tool to evaluate now — especially valuable for organizations that have been waiting for enterprise-grade AI security tooling rather than stitching together point solutions. The key question for enterprise security teams is integration: does this slot into your existing SIEM and vulnerability management workflow, or does it create another silo? Glasswing's open-source support angle also matters — if you're a security team maintaining or depending on open-source infrastructure, Anthropic's research-sharing commitment could be practically valuable. Evaluate the beta; the window for shaping your AI security stack before it's set is narrow.
7. Google's Post-I/O Fallout: Gemini Agents, DuckDuckGo Backlash, and the New Rules of AI Search Visibility
The aftershocks from Google I/O 2026 continue to ripple through the search and marketing ecosystem. Google's "agentic era" launch — which included Gemini 3.5 Flash, the Gemini Spark personal agent, a redesigned AI-first Search interface capable of processing text, images, video, and browser tabs, and the introduction of Android Halo for tracking agent task progress — is reshaping how businesses think about digital visibility.
Google's I/O blog post described 100 announcements in total, including "information agents" that continuously monitor topics, news, and shopping trends on users' behalf. This is the critical shift: search is evolving from a pull model (user queries, Google retrieves) to a push model (agents proactively surface information before users ask). For businesses that depend on being found at the moment of intent, this fundamentally changes the competitive dynamics.
At the same time, Google's "Preferred Sources" feature — announced this week — attempts to build a bridge for publishers. Any website publishing fresh, relevant content is now eligible to appear as a preferred source inside AI Overviews and AI Mode results. This is Google's answer to the publisher revolt over AI summaries cannibalizing traffic: you can still get visibility, but the mechanism is now AI curation rather than ranking.
"Google's new 'AI Mode' shifts discovery from search rankings to AI-generated recommendations, changing how businesses gain visibility online." — PYMNTS
The Google I/O aftermath is where AI strategy meets revenue reality for most businesses. The core insight: SEO is no longer primarily about ranking — it's about being cited. AI agents don't rank your page; they either reference your content as an authoritative source or they don't. That means the investment in E-E-A-T (experience, expertise, authoritativeness, trustworthiness) content — high-quality, frequently updated, clearly attributed — is more valuable than ever. Generic, SEO-farmed content will be invisible in an agent-mediated search world. The winners in AI search will be brands that have built genuine topical authority, not just keyword coverage. Start auditing your content library for citation-worthiness, not just ranking position.
Why It All Matters — The Week's Throughline
This week's stories connect through a single thread: AI accountability is going mainstream, but it's arriving unevenly. States are stepping up where federal regulators won't. Companies are building safety narratives that may or may not track with their actual practices. Search — the infrastructure of online business — is being rewired in ways that disadvantage businesses that haven't adapted their content strategy. And the AI jobs debate has shifted from apocalyptic to nuanced, which is more useful but also harder to act on.
The common denominator for enterprise decision-makers: stop waiting for consensus and start making specific, defensible choices. Which AI vendors can demonstrate independent safety verification? What does your digital visibility strategy look like in an agent-mediated search world? Which specific job functions in your organization are genuinely at automation risk right now — not in a theoretical future? These aren't abstract questions anymore. The regulatory, market, and competitive signals are all pointing to the same conclusion: the businesses that thrive in the AI era will be the ones that got granular early.
Navigating AI strategy while the rules keep changing?
SEN-X helps enterprise teams turn these developments into concrete, actionable roadmaps — from AI search strategy to vendor due diligence to workforce planning.
Contact SEN-X →