Skip to navigation

Threat reportThreat IntelligenceTL-2026-3055

OpenAI disrupts Russian 'Dark Clark' fake-think-tank and Iranian fake-journalist AI-enabled influence operations

mediumMONITORING

OpenAI disclosed takedowns of two AI-enabled false-front influence operations. A Russian-origin cluster ('Dark Clark') ran a fake Latin American think tank, the Social Research Center, fronted by an AI persona 'Mia Clark', to target Latin American politics and damage Ukraine's reputation, including fake audio in Ecuador and Bolivia.

Key facts for Dark Clark influence operation

IdentifiersTL-2026-3055
Also known asDark Clark, Bogus Bylines
SeverityMEDIUM
First disclosed
Exploitation statusNo known exploitation
Affected productsVarious Online news and international-affairs outlets; social platforms (TikTok, Facebook, X)
AttributionDark Clark (low confidence)
Nation-state nexusRussia, Iran
ToolsChatGPT
MITRE ATT&CK techniquesT1565.001 Stored Data Manipulation, T1583.006 Web Services, T1585.001 Social Media Accounts, T1588.007 Artificial Intelligence, T1684.001 Impersonation
MotivationUNKNOWN
Target sectorsnews - media, government administration, think tanks, civil society
Target regionsLatin America, argentina, bolivia, ecuador, united states of america, Middle East
Detection rules9 (1 shown free)
Sources7 cited sources

Detection coverage for Dark Clark influence operation

As of 2026-10-08, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-3055 across Splunk SPL, Microsoft KQL and Sigma, covering 13 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.

Free detection: Inbound Mail From Known Fake-Journalist / Front-Identity Personas (Dark Clark, Iranian Bylines)

Flags inbound email whose sender address or subject carries a persona named in the OpenAI false-front report: the Russian-origin 'Mia Clark' / Social Research Center front or one of the seven Iranian-linked fake journalists (Ervin B. Hoskins, Noah Lamington, Ericka Feusier, plus the common-name personas Sophia Gonzalez, Michael Harrison, Jenny Williams, Alice Johnson). Common-name personas only alert when paired with op-ed / pitch / Iran-geopolitics subject language to limit noise. Aimed at newsroom, editorial and comms mail flows where unsolicited article pitches arrive. Persona names come partly from secondary coverage; this is entity matching, not malware detection.

  • Languages: Splunk SPL, Microsoft KQL, Sigma
  • ATT&CK: T1585.001, T1585.002, T1656
  • Data sources: o365_message_trace, defender_emailevents
  • Severity: medium
  • Confidence: medium

Splunk SPL

index=o365 sourcetype="o365:reporting:messagetrace" earliest=-24h
| eval sender=lower(SenderAddress), subj=lower(Subject)
| eval hi_spec=if(match(sender,"(hoskins|lamington|feusier|mia[._-]?clark|socialresearchcent)") OR match(subj,"(ervin b[.]? hoskins|noah lamington|ericka feusier|mia clark|social research cent(er|re))"),1,0)
| eval common=if(match(sender,"(sophia[._-]?gonzalez|michael[._-]?harrison|jenny[._-]?williams|alice[._-]?johnson)") AND match(subj,"(iran|tehran|middle east|geopolit|op-?ed|pitch|contribut|guest (article|post))"),1,0)
| where hi_spec=1 OR common=1
| stats count AS messages min(_time) AS first_seen max(_time) AS last_seen values(Subject) AS subjects values(RecipientAddress) AS recipients BY SenderAddress
| convert ctime(first_seen) ctime(last_seen)
| sort - messages

Microsoft KQL

// Inbound mail from known fake-journalist / front-identity personas
EmailEvents
| where Timestamp > ago(24h)
| where EmailDirection == "Inbound"
| extend sender = tolower(SenderFromAddress), disp = tolower(SenderDisplayName), subj = tolower(Subject)
| extend HiSpec = (disp matches regex @"(ervin b[.]? hoskins|noah lamington|ericka feusier|mia clark)")
    or (sender matches regex @"(hoskins|lamington|feusier|mia[._-]?clark|socialresearchcent)")
    or (subj matches regex @"(social research cent(er|re))")
| extend Common = ((disp matches regex @"(sophia gonzalez|michael harrison|jenny williams|alice johnson)")
    or (sender matches regex @"(sophia[._-]?gonzalez|michael[._-]?harrison|jenny[._-]?williams|alice[._-]?johnson)"))
    and (subj matches regex @"(iran|tehran|middle east|geopolit|op-?ed|pitch|contribut|guest (article|post))")
| where HiSpec or Common
| project Timestamp, NetworkMessageId, SenderFromAddress, SenderDisplayName, RecipientEmailAddress, Subject, DeliveryAction, AuthenticationDetails, HiSpec, Common

Sigma

title: Inbound Email From Known Fake-Journalist or Front-Identity Personas
id: 7c1e3a52-4b0d-4e6a-9f1b-2a5d8c3e1f01
status: experimental
description: Detects inbound email whose sender display name, sender address or subject matches personas named in the OpenAI false-front operations report (Mia Clark / Social Research Center; Iranian-linked fake journalists). Common-name personas only match with pitch or Iran-geopolitics subject terms.
references:
    - https://cyberscoop.com/openai-disrupts-russia-iran-ai-influence-operations/
    - https://openai.com/index/disrupting-ai-enabled-false-front-operations/
author: AII-Detector — ThreadLinqs Intelligence
date: 2026-10-08
tags:
    - attack.resource-development
    - attack.t1585.001
    - attack.t1585.002
    - attack.stealth
    - attack.t1656
logsource:
    product: microsoft365defender
    service: emailevents
detection:
    selection_hi_name:
        SenderDisplayName|contains:
            - 'Ervin B. Hoskins'
            - 'Ervin Hoskins'
            - 'Noah Lamington'
            - 'Ericka Feusier'
            - 'Mia Clark'
    selection_hi_addr:
        SenderFromAddress|contains:
            - 'hoskins'
            - 'lamington'
            - 'feusier'
            - 'mia.clark'
            - 'mia_clark'
            - 'miaclark'
            - 'socialresearchcent'
    selection_common:
        SenderDisplayName|contains:
            - 'Sophia Gonzalez'
            - 'Michael Harrison'
            - 'Jenny Williams'
            - 'Alice Johnson'
    selection_topic:
        Subject|contains:
            - 'Iran'
            - 'Tehran'
            - 'Middle East'
            - 'geopolit'
            - 'op-ed'
            - 'pitch'
            - 'guest article'
            - 'guest post'
    condition: selection_hi_name or selection_hi_addr or (selection_common and selection_topic)
falsepositives:
    - Real people who share these common names writing to the organisation
    - Legitimate freelancers pitching Iran or Middle East coverage
    - The real Social Research Centre (Australia) corresponding with partners
level: medium

Known false positives:

  • Real individuals who share the common persona names (Michael Harrison, Jenny Williams, Alice Johnson, Sophia Gonzalez) sending genuine pitches
  • Legitimate freelance journalists pitching Iran or Middle East coverage

Tuning notes: Pitches from the personas arrive from throwaway mailboxes (T1585.002), so also review SPF/DKIM/DMARC results and sender-domain age for hits. Treat a hit as a lead for editorial verification, not as proof: check byline history and outlet cross-references before acting. The high-specificity names (Hoskins, Lamington, Feusier, Mia Clark) are the reliable triggers; drop the common-name branch if it is noisy. Persona names are partly from secondary coverage, so refresh the list when OpenAI's primary report or other vendors publish more.

8 more detection rules (Splunk SPL, Microsoft KQL, Sigma) · unlock with TL Creds. Sign up free to unlock the detection rules or query them with the Threadlinqs MCP server (Purple)
  • Inbound Email Linking To Social Research Center Front Infrastructure · unlock with TL Creds
  • Entra ID Account Created Or Invited Using A Flagged Persona Name · unlock with TL Creds
  • LLM Gateway Prompts Referencing Dark Clark Front Identity Or Fake-Byline Pitching · unlock with TL Creds
  • Audio Attachment From Official-Sounding External Sender (Fake Official Audio) · unlock with TL Creds
  • File Created With Flagged Persona Or Front-Organization Name · unlock with TL Creds
  • Proxy Request For Flagged Fake-Journalist Persona Profile URL · unlock with TL Creds
  • Fake-Journalist Byline / Author Pages Served By Your Outlet · unlock with TL Creds
  • DNS Lookups For Social Research Center / Mia Clark Front-Site Hostnames · unlock with TL Creds

Indicators of compromise for Dark Clark influence operation

As of 2026-10-08, Threadlinqs Intelligence tracks 13 indicators of compromise for TL-2026-3055.

13 indicators of compromise · unlock with TL Creds. Compare plans

How Dark Clark influence operation works

An Iranian-linked, for-hire-style campaign used at least seven fake journalist personas to place almost 100 articles across about a dozen outlets on the U.S.-Iran war.

On 2026-10-08 OpenAI published 'Disrupting AI-enabled false front operations' describing two covert influence operations (IO) that used ChatGPT and other AI tools to build false-front identities: a fake think tank and fake journalists. OpenAI rated the Russian operation 5 and the Iranian operation 4 on its 6-point Breakout Scale (most campaigns it sees rate 1 or 2). CyberScoop reports it as the first time OpenAI has reported a high-impact influence campaign.

Russian-origin operation 'Dark Clark': The cluster is named for 'Mia Clark', a fabricated AI persona presented as leader of the Social Research Center (SRC), a purported Latin American think tank. Its focus was Latin American politics and culture, mainly harming Ukraine's reputation, with engagement on political issues in Argentina and Bolivia (secondary coverage also lists Ecuador, Peru and Poland and reports more than 60 original articles). OpenAI assessed that the Russian group controlled the SRC, citing ChatGPT-generated internal reports on staff wage scales and hiring and firing, and said the SRC's Latin American staff were not aware they worked for a Russian group. OpenAI called it the most complex front-identity attempt it has disrupted in two and a half years. In March a TikTok video falsely claiming the Ecuadorian government used shell companies to recruit citizens to fight for Ukraine was fact-checked by Lupa Media, and the operation also spread fake audio of the Ukrainian consul in Ecuador insulting Ecuadorians. During May anti-government protests in Bolivia it released fake audio of a state water company employee claiming a state of emergency and water cuts for La Paz; the Bolivian government denied it on Facebook. Operators also queried the model about Politology, described as a successor to the Wagner Group, far more than any other Russian organization. No specific Russian government or intelligence agency was named.

Iranian operation: Using many of the same AI tools, the operation pitched and placed articles on the U.S.-Iran war under at least seven fake journalist personas. OpenAI identified almost 100 articles published or syndicated under those bylines across about a dozen small-to-medium international-affairs, geopolitics and Middle East outlets, and the operation also generated social media comments. Secondary coverage rates article placement Category 4 and social media impact Category 2. Most stories drew little social engagement. OpenAI said the activity resembled a commercial actor running a for-hire campaign and did not attribute it to a specific actor or group. Both operations reportedly overstated their own effectiveness in internal reports. The number of banned accounts was not stated. No CVEs, malware or network IOCs were published; indicators are persona and organization names. Some persona names come from secondary coverage of the OpenAI report because the primary page was not retrievable.

MITRE ATT&CK techniques by tactic:

Impact

T1565.001 Stored Data Manipulation

Resource Development

T1583.006 Web Services; T1585.001 Social Media Accounts; T1588.007 Artificial Intelligence

Stealth

T1684.001 Impersonation

Remediation for Dark Clark influence operation

Immediate actions

  • Treat the named personas and the Social Research Center as untrusted sources; check whether they appear in your media-monitoring or sourcing lists
  • Verify audio and video of officials and utility employees through the originating government or company channel before amplifying

Workarounds

  • Require editorial verification of unsolicited op-ed pitches from unknown contributors

Longer-term hardening

  • Add provenance checks for think-tank and freelance-journalist submissions (identity verification, byline history, outlet cross-checks)
  • Track AI-enabled influence operation reporting and share persona indicators with fact-checking networks

Affected and fixed versions

  • Various — Online news and international-affairs outlets; social platforms (TikTok, Facebook, X)

Timeline of Dark Clark influence operation

  • Operation also spreads fake audio of the Ukrainian consul in Ecuador insulting Ecuadorians (timing not stated; placed here with the Ecuador activity)
  • A locked-ioc-13 video falsely claiming the Ecuadorian government used shell companies to recruit citizens to fight for Ukraine is fact-checked by Lupa Media (month precision; exact day not stated)
  • Bolivian government publicly denies the fake audio on locked-ioc-11 (month precision)
  • During anti-government protests, fake audio of a state water company employee claims a state of emergency and water cuts for La Paz (month precision)
  • CyberScoop reports the takedowns, noting OpenAI's first report of a high-impact influence campaign
  • OpenAI publishes its report on the disrupted Russian 'Dark Clark' (Breakout Scale 5) and Iranian fake-journalist (Breakout Scale 4) operations

Sources cited for Dark Clark influence operation

Previous and next in threat intel

Drafted by the Threadlinqs research pipeline from 7 sources; published ; last updated —.

Threadlinqs Intelligence — Real-Time Threat Detection Platform

[ 0 threats ] [ 0 det ] [ CRIT: 0 ] [ HIGH: 0 ]
// threat_feed
$ sort --newest
Showing all threats

Live intelligence console

Threat level

Latest Threats