Daily AI intelligence for Iru.
Wednesday, May 27, 2026
  • Gemini 3.5 Flash hits general availability and now powers AI Mode Search for over one billion users.
  • Fireworks AI is in talks to raise at a $15 billion valuation, nearly 4x its last known mark.
  • Ping Identity ships an AI agent governance layer that manages non-human identities across their full lifecycle.
  • CrowdStrike and Google jointly dismantled the Glassworm botnet, which had infected hundreds of open-source packages since early 2025.
  • Anthropic overtakes OpenAI in enterprise adoption share for the first time, per Ramp's May AI Index.
Gemini 3.5 Flash Reaches General Availability
aitoolsrecap.com · May 27

Google's Gemini 3.5 Flash moved to general availability on May 27, becoming the inference backbone for AI Mode Search across its billion-user base.

  • Scale signal: Powering AI Mode Search at this user count makes Gemini 3.5 Flash the highest-traffic frontier model deployment announced to date.
  • Enterprise relevance: GA status means Gemini 3.5 Flash is now available with production SLAs for API customers, removing the preview caveat that blocked procurement sign-off.
Bottom line
Google now has a GA frontier model at consumer scale, compressing the evaluation timeline for enterprise teams still benchmarking against GPT-5 and Claude.
Ping Identity Adds Agent Governance to Identity Platform
prnewswire.com · May 27

Ping Identity shipped new platform capabilities on May 27 that treat AI agents as first-class identity principals, giving enterprises discovery, lifecycle governance, and secretless privileged access for non-human actors.

  • What's new: AI-first headless interfaces let developers register and govern agents programmatically; privileged access controls let agents operate without exposing credentials.
  • Why it matters: Most IAM platforms were built assuming human actors; Ping is the first major incumbent to ship a dedicated lifecycle model for agents rather than retrofitting service accounts.
Bottom line
With AI agents multiplying faster than IT teams can track them, identity governance for non-human actors is becoming the enterprise security gap that IAM vendors are racing to close.
Anthropic Ships Free Security Plugin for Claude Code
cybersecuritynews.com · May 27

Anthropic released a free security-guidance plugin for its Claude Code terminal tool that reviews code edits, model outputs, and commits in real time to catch vulnerabilities before they reach production.

  • Mechanism: The plugin runs autonomously within the Claude Code terminal session, flagging issues inline without requiring a separate SAST or SCA pipeline call.
  • Positioning: Releasing it free shifts the competitive frame — AppSec tooling becomes a Claude Code retention feature rather than a standalone revenue line.
Bottom line
Anthropic is embedding security review into the AI coding workflow itself, compressing the feedback loop that traditionally required a separate security gate.
Slack Workflow Builder Gains Native AI Response Step
uctoday.com · May 27

Slack launched a Generate AI Response step inside Workflow Builder on May 27, letting non-technical employees inject summarization, translation, and drafting into automated business processes without code.

  • Scope: The step is a native workflow block, meaning it composes with existing triggers and actions rather than requiring a separate AI integration or API key.
  • Competitive context: Microsoft Copilot Studio has offered comparable no-code AI workflow composition for over a year; this closes a visible gap for Slack-native shops.
Bottom line
Slack is turning AI from a chat-tab feature into a process automation primitive, which changes how IT and ops teams should evaluate its place in the enterprise stack.

Fireworks AI, which sells managed inference infrastructure for enterprise model deployments, is in talks to raise at a $15 billion valuation co-led by Index Ventures — nearly four times its last known $4 billion mark — signaling that inference-layer infrastructure commands frontier-model multiples even without owning the models.

Fujitsu announced a strategic partnership with Anthropic on May 27 to commercialize autonomous multi-agent systems that improve themselves between sessions, betting that the next enterprise baseline is AI that removes the human-in-the-loop requirement rather than accommodating it.

CrowdStrike and Google Take Down Glassworm Supply-Chain Botnet
cyberscoop.com · May 27

CrowdStrike, Google, and Shadowserver simultaneously seized four command-and-control servers on May 27, dismantling the Glassworm botnet that had been injecting malware into hundreds of open-source packages since early 2025.

  • Scale of compromise: Hundreds of open-source software packages were infected, meaning any enterprise consuming public package registries without provenance controls had attack-surface exposure.
  • Method: The simultaneous multi-server takedown was designed to defeat the botnet's resilience architecture, which was engineered to survive single-node disruptions.
Bottom line
Glassworm's 18-month run inside the open-source supply chain is the clearest recent evidence that software composition analysis and package provenance verification are non-optional controls for enterprise dev pipelines.
  • Anthropic surpasses OpenAI in enterprise adoption Ramp's May AI Index showing Anthropic at 34.4% of paying businesses versus OpenAI at 32.3% is generating debate about whether this reflects Claude's coding performance, enterprise trust posture, or API pricing — and whether it signals a durable shift or a snapshot.
  • Token budgets becoming enterprise cost crisis Ethan Mollick's observation that companies swing between 'use all the tokens' and 'use local models' with no coherent strategy in between is resonating widely, with practitioners sharing that token spend materialized as a budget line almost overnight and no one has governance processes to match.
  • Grok content review publicly called broken X product head Nikita Bier's admission that the Grok automated content-review system is 'broken' after a wave of false adult-content flags on harmless posts is drawing attention to the risks of deploying AI moderation at platform scale without adequate feedback loops.
  • AI IPO wave drawing dot-com comparisons Commentary circulating on X argues that the combined OpenAI, Anthropic, and SpaceX IPOs could raise as much capital as the entire 2000 dot-com IPO class, prompting debate about whether the analogy signals peak euphoria or a structurally different demand environment.
  • Codex scope expanding beyond coding OpenAI's Greg Brockman posted that Codex is useful for 'any kind of work done with a computer,' not just software development, triggering discussion about whether the agent framing has quietly outgrown its developer-tool positioning.

← Back to latest