Glow says AI coding agents asked for review screenshots put over 13,000 internal images from 300-plus organizations into public GitHub repositories. Most sat under developers' personal accounts, where the companies' security teams were not looking.
Perspective Coverage
5 publishers
- Builder
- Builder 37%
- Operator
- Operator 53%
- Investor
- Investor 10%
Reality
- Evidence55
- Adoption45
- Hype gap+15
- Incentives70
- Confidence60
Delinea's 2026 survey found 99.7% of IT and security leaders have a formal AI data policy, but only about 51% check AI access against it in real time. For responders, the shortfall comes after an agent starts acting: how fast out-of-scope access is seen, and how fast it is taken away.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+20
- Incentives60
- Confidence40
Glow's PixelLeak report found coding agents exposed more than 13,000 private screenshots from over 300 organisations by hosting them in public GitHub repos. With 93% under developers' personal accounts, a company's own GitHub audit would miss most of them.
Perspective Coverage
3 publishers
- Builder
- Builder 42%
- Operator
- Operator 50%
- Investor
- Investor 8%
Reality
- Evidence55
- Adoption50
- Hype gap+10
- Incentives55
- Confidence60
BlackFog's ADX Vision 2.0 runs seven layers of prompt-injection checks on the endpoint, covering prompts employees write and prompts AI agents send. Teams comparing it with per-vendor AI controls have only the company's own description to go on.
Reality
- Evidence22
- Adoption
- Insufficient
- Hype gap+35
- Incentives75
- Confidence30
Detectify found 86% to 97% of open critical and high flaws on 1,293 customers' internet-facing systems had been exposed for more than 90 days. Its scanner confirmed each with a working attack request, so these are known, reachable flaws left to age.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+10
- Incentives60
- Confidence55
LastPass added shadow AI discovery and sensitive-data warnings to Business Max, run from the browser extension its customers already use. Teams on that tier get a warning and an audit log for browser AI use, and the release describes hard blocks only for web categories.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+25
- Incentives75
- Confidence45
Omnissa launched Elara for AI governance, citing its own report that AI-assistant use on work devices grew nearly 1,000% in 2025, mostly via unapproved tools. For endpoint teams it is a new category to assess, and the controls Omnissa described attach most clearly to agents IT has already granted access.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+40
- Incentives75
- Confidence40
AI token caps are pushing accountants onto free chatbots as the Oct. 15 extension deadline nears, CPA Practice Advisor reports. The overage a firm saves is known in advance, while the cost of client records held by an unvetted vendor is set only after a breach.
Reality
- Evidence38
- Adoption
- Insufficient
- Hype gap+30
- Incentives62
- Confidence42
SOCRadar tied 5,434 infostealer records for AI tools to 1,500 corporate email addresses at 482 large enterprises. The report says those AI accounts belong under the same sign-on and session controls as a company's identity provider and code repositories.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+35
- Incentives85
- Confidence40
SOCRadar found captured ChatGPT sessions at 358 of the 482 companies whose AI accounts turned up in 90 days of infostealer logs. The firm ties the spread to shadow AI, with employees opening work-email accounts that IT never sees.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+15
- Incentives
- Insufficient
- Confidence45
AWS's Reimagine 2026 report, built on 154 executive interviews, says review processes sized for six-month IT programs are driving AI use underground. AWS's answer is governance built into the systems themselves, with humans kept accountable for outcomes.
Reality
- Evidence40
- Adoption25
- Hype gap+10
- Incentives65
- Confidence45
Okta and 11 vendors including AWS, Google Cloud and CrowdStrike signed six shared principles for securing AI agents under a new Blueprint Alliance. Buyers can put the list to vendors in procurement now, while Okta's release describes a reference architecture with no stated way to test compliance.
Publishers:okta.com · scworld.com Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+35
- Incentives70
- Confidence55
Only 36% of IT staff in an Automox survey feel highly confident in their endpoint compliance visibility, as AI agents start acting on managed devices. Agents inherit their launcher's rights, so teams now need a record of what each may do and what it did.
Reality
- Evidence40
- Adoption50
- Hype gap+25
- Incentives80
- Confidence40
Microsoft's Purview and Entra Global Secure Access can now block sensitive uploads to unsanctioned AI tools at the network layer, for users and agents. Licensing is not in the announcement, so tenants cannot yet tell whether it is a setting or a purchase.
Reality
- Evidence35
- Adoption
- Insufficient
- Hype gap+20
- Incentives85
- Confidence40
An IANS and Artico survey puts AI at about 3 percent of the average security budget, and 24 percent of organizations have given it a line of its own. Bugcrowd's CEO says the rest comes out of what was already bought.
Reality
- Evidence40
- Adoption48
- Hype gap+15
- Incentives78
- Confidence45
A Microsoft education partner argues the evaluation window is itself the exposure window, citing a survey in which most security professionals knew or suspected unapproved AI use. The test he proposes is one conversation with the CIO or CISO and four document requests.
Reality
- Evidence42
- Adoption61
- Hype gap+30
- Incentives82
- Confidence41
The general availability release arrives with Darktrace's own sensor data as its evidence. That data shows how widely AI gets used inside customer estates; the unapproved share rests on three deployments Darktrace did not name.
Reality
- Evidence38
- Adoption34
- Hype gap+18
- Incentives74
- Confidence58
Agent Gateway enforces policy on each tool call and logs it as it happens, while deactivating an agent in the console today only blocks new sessions. Okta says revoking live tokens and cutting sessions in flight is planned.
Reality
- Evidence42
- Adoption27
- Hype gap+30
- Incentives80
- Confidence55
Agent SSO went generally available on August 24 at no extra cost inside core Okta SSO, registering Cross App Access agents in Universal Directory and issuing short-lived tokens. Discovering unregistered agents takes a separate subscription.
Publishers:okta.com
Reality
- Evidence34
- Adoption22
- Hype gap+35
- Incentives88
- Confidence58
The third wave of the Konstanz AI study surveyed 1,105 employees in May 2026 and put workplace AI use at 38 percent, a three-point rise that sits inside the sample's own error bar. Just over half of the tools those users reach for came in through the employer.
Reality
- Evidence60
- Adoption45
- Hype gap+15
- Incentives40
- Confidence55
Earlier coverage
- Deloitte puts UK workers' own spending on work GenAI tools at £958m a year
Leadership · September 18, 2026 · 1 publisher
- Zscaler's Rob Sloan makes California's agency rule the case for a corporate AI security owner
Leadership · September 18, 2026 · 1 publisher
- Nearly half of surveyed organizations had an AI agent take an unapproved action in the past year
Security · September 15, 2026 · 1 publisher
- Exaforce ties every AI agent it discovers back to the employee whose permissions it borrows
Product · September 15, 2026 · 1 publisher
- A personal AI account inherits consumer terms and everything its user can reach
Build · September 15, 2026 · 1 publisher
- Team8's survey finds 71% of CISOs already putting AI agents into their security tools
Security · September 14, 2026 · 1 publisher
- Weekly AI tax research nearly doubled to 60% among more than 1,000 tax professionals
Invest · September 12, 2026 · 1 publisher
- Amazon's cure for shadow AI ships on the desktop from outside its European sovereign cloud
Product · September 12, 2026 · 1 publisher
- Alvarez & Marsal's talent practice lead warns AI pilots often test the wrong thing, and premature cuts create talent debt
Leadership · September 11, 2026 · 1 publisher
- Hunt.io traces intrusions in four countries to AI agents running eight known exploits
Security · September 10, 2026 · 1 publisher
- Amazon pitches the Quick desktop app to IT as containment for shadow AI
Build · September 10, 2026 · 1 publisher
- Copilot Chat read confidential mail for weeks past the DLP policy set to stop it
Leadership · September 10, 2026 · 1 publisher
- OpenAI test found agents breaching Hugging Face; CrowdStrike touts new tools to police shadow AI
Product · September 10, 2026 · 1 publisher
- Uber exhausted a year of Claude Code budget in four months
Leadership · September 9, 2026 · 1 publisher
- Orphaned AI agents keep their access after the employee who created them moves on
Product · September 7, 2026 · 1 publisher
- Anthropic's Compliance API now logs the Claude Code sessions running on inherited developer credentials
Security · August 31, 2026 · 1 publisher
- Six to nine vendors, five obligations each: the first AI security exercise is arithmetic
Build · August 26, 2026 · 1 publisher
- Akamai's 12x number turns shadow AI from a headcount problem into a shortlist
Security · August 24, 2026 · 1 publisher
- IT's AI shopping list is inverted: 46.5% want automation, 71% of their AI tools are invisible
Product · August 22, 2026 · 1 publisher
- AI risk filed in the wrong drawer: agents hold staff entitlements and nobody signs for them
Security · August 22, 2026 · 1 publisher
- Approval is a snapshot: the same sanctioned app becomes shadow AI 24 minutes later
Security · August 22, 2026 · 1 publisher
- Shadow AI now has an invoice: about $670K on top of the average breach
Build · August 16, 2026 · 1 publisher