Vulnerability-Lookup

CVE-2026-17351 (GCVE-0-2026-17351)

Vulnerability from cvelistv5 – Published: 2026-07-31 16:00 – Updated: 2026-08-01 03:56

Title

pgAdmin 4: AI Assistant read-only transaction bypass via sqlparse/PostgreSQL lexer disagreement (incomplete fix for CVE-2026-12045)

Summary

The fix for CVE-2026-12045 in pgAdmin 4 9.16 required the LLM-supplied query passed to the AI Assistant's execute_sql_query tool to parse, via sqlparse, as exactly one non-transaction-control statement before running it inside a BEGIN TRANSACTION READ ONLY wrapper. sqlparse's string-literal lexing can disagree with PostgreSQL's own parser: under standard_conforming_strings = on (PostgreSQL's default since 9.1), a backslash immediately before a quote is an ordinary character to PostgreSQL, but sqlparse treats it as escaping the quote. A payload such as SELECT '\';COMMIT;CREATE TABLE pwn(x int);SELECT 1 --' therefore parses as a single SELECT to sqlparse's validator, while PostgreSQL executes it as four statements: the smuggled COMMIT ends the wrapping read-only transaction, and the trailing ROLLBACK becomes a no-op. This reintroduces the same write/RCE bypass CVE-2026-12045 was meant to close, reachable via the same indirect prompt-injection delivery (an attacker plants the payload in any object the AI Assistant may read; the LLM emits it as a tool call). An initial candidate fix ran the query with psycopg's execute(..., prepare=True), intending to force PostgreSQL's own Parse step (extended query protocol) to reject multi-statement text regardless of sqlparse's classification. This candidate fix does not work as submitted: psycopg3's PrepareManager silently ignores the prepare argument whenever the connection's prepare_threshold is None, which is pgAdmin's default for every server connection (the per-server "Prepare threshold" field is blank unless an administrator explicitly sets it) -- psycopg3 falls back to the simple query protocol, the same multi-statement-capable path the bypass exploits, so the candidate fix closes nothing on any real-world default configuration. The corrected fix sets conn.prepare_threshold = 0 directly on the dedicated, single-use read-only connection the AI Assistant tool opens, structurally forcing the extended query protocol independent of any server-level configuration. Verified against a live PostgreSQL 18 instance: the payload executes successfully under the prepare_threshold=None (default) behavior, and is rejected with "cannot insert multiple commands into a prepared statement" once prepare_threshold=0 is set on that connection. This issue affects pgAdmin 4: from 9.13 before 9.17.

Severity

9 (Critical)


                        
                          CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H

9.4 (Critical)


                        
                          CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H

SSVC

Exploitation: none Automatable: no Technical Impact: total

CISA Coordinator (v2.0.3)

CWE

CWE-89 - Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection')
CWE-115 - Misinterpretation of Input

Assigner

PostgreSQL

References

3 references

URL	Tags
https://github.com/pgadmin-org/pgadmin4/issues/10192	issue-tracking
https://github.com/pgadmin-org/pgadmin4/commit/ef…	patch
https://github.com/pgadmin-org/pgadmin4/commit/bf…	related

Impacted products

1 product

Vendor	Product	Version
pgadmin.org	pgAdmin 4	Affected: 9.13 , < 9.17 (custom)

Credits

Kai Aizen (SnailSploit) Kundan Sable <kundan.sable@enterprisedb.com> Ashesh Vashi <ashesh.vashi@enterprisedb.com> Ashesh Vashi <ashesh.vashi@enterprisedb.com>

Show details on NVD website

JSON

To clipboard

{
  "containers": {
    "adp": [
      {
        "metrics": [
          {
            "other": {
              "content": {
                "id": "CVE-2026-17351",
                "options": [
                  {
                    "Exploitation": "none"
                  },
                  {
                    "Automatable": "no"
                  },
                  {
                    "Technical Impact": "total"
                  }
                ],
                "role": "CISA Coordinator",
                "timestamp": "2026-07-31T00:00:00+00:00",
                "version": "2.0.3"
              },
              "type": "ssvc"
            }
          }
        ],
        "providerMetadata": {
          "dateUpdated": "2026-08-01T03:56:14.973Z",
          "orgId": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
          "shortName": "CISA-ADP"
        },
        "title": "CISA ADP Vulnrichment"
      }
    ],
    "cna": {
      "affected": [
        {
          "defaultStatus": "unaffected",
          "modules": [
            "AI Assistant",
            "LLM Tools"
          ],
          "product": "pgAdmin 4",
          "programFiles": [
            "https://github.com/pgadmin-org/pgadmin4/blob/master/web/pgadmin/llm/tools/database.py"
          ],
          "repo": "https://github.com/pgadmin-org/pgadmin4",
          "vendor": "pgadmin.org",
          "versions": [
            {
              "lessThan": "9.17",
              "status": "affected",
              "version": "9.13",
              "versionType": "custom"
            }
          ]
        }
      ],
      "credits": [
        {
          "lang": "en",
          "type": "finder",
          "value": "Kai Aizen (SnailSploit)"
        },
        {
          "lang": "en",
          "type": "remediation developer",
          "value": "Kundan Sable \u003ckundan.sable@enterprisedb.com\u003e"
        },
        {
          "lang": "en",
          "type": "remediation developer",
          "value": "Ashesh Vashi \u003cashesh.vashi@enterprisedb.com\u003e"
        },
        {
          "lang": "en",
          "type": "remediation reviewer",
          "value": "Ashesh Vashi \u003cashesh.vashi@enterprisedb.com\u003e"
        }
      ],
      "descriptions": [
        {
          "lang": "en",
          "value": "The fix for CVE-2026-12045 in pgAdmin 4 9.16 required the LLM-supplied query passed to the AI Assistant\u0027s execute_sql_query tool to parse, via sqlparse, as exactly one non-transaction-control statement before running it inside a BEGIN TRANSACTION READ ONLY wrapper. sqlparse\u0027s string-literal lexing can disagree with PostgreSQL\u0027s own parser: under standard_conforming_strings = on (PostgreSQL\u0027s default since 9.1), a backslash immediately before a quote is an ordinary character to PostgreSQL, but sqlparse treats it as escaping the quote. A payload such as SELECT \u0027\\\u0027;COMMIT;CREATE TABLE pwn(x int);SELECT 1 --\u0027 therefore parses as a single SELECT to sqlparse\u0027s validator, while PostgreSQL executes it as four statements: the smuggled COMMIT ends the wrapping read-only transaction, and the trailing ROLLBACK becomes a no-op. This reintroduces the same write/RCE bypass CVE-2026-12045 was meant to close, reachable via the same indirect prompt-injection delivery (an attacker plants the payload in any object the AI Assistant may read; the LLM emits it as a tool call).\n\nAn initial candidate fix ran the query with psycopg\u0027s execute(..., prepare=True), intending to force PostgreSQL\u0027s own Parse step (extended query protocol) to reject multi-statement text regardless of sqlparse\u0027s classification. This candidate fix does not work as submitted: psycopg3\u0027s PrepareManager silently ignores the prepare argument whenever the connection\u0027s prepare_threshold is None, which is pgAdmin\u0027s default for every server connection (the per-server \"Prepare threshold\" field is blank unless an administrator explicitly sets it) -- psycopg3 falls back to the simple query protocol, the same multi-statement-capable path the bypass exploits, so the candidate fix closes nothing on any real-world default configuration.\n\nThe corrected fix sets conn.prepare_threshold = 0 directly on the dedicated, single-use read-only connection the AI Assistant tool opens, structurally forcing the extended query protocol independent of any server-level configuration. Verified against a live PostgreSQL 18 instance: the payload executes successfully under the prepare_threshold=None (default) behavior, and is rejected with \"cannot insert multiple commands into a prepared statement\" once prepare_threshold=0 is set on that connection.\n\nThis issue affects pgAdmin 4: from 9.13 before 9.17."
        }
      ],
      "metrics": [
        {
          "cvssV3_1": {
            "attackComplexity": "LOW",
            "attackVector": "NETWORK",
            "availabilityImpact": "HIGH",
            "baseScore": 9,
            "baseSeverity": "CRITICAL",
            "confidentialityImpact": "HIGH",
            "integrityImpact": "HIGH",
            "privilegesRequired": "LOW",
            "scope": "CHANGED",
            "userInteraction": "REQUIRED",
            "vectorString": "CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H",
            "version": "3.1"
          },
          "format": "CVSS",
          "scenarios": [
            {
              "lang": "en",
              "value": "Threat model: identical to CVE-2026-12045 -- indirect prompt injection. An attacker with low-privilege DB write access plants a payload in any object the AI Assistant might read (table row, column value, comment). When a pgAdmin user with a higher-privilege database role (typically superuser or holder of pg_execute_server_program) opens the AI Assistant against that content, the LLM emits the lexer-differential payload as a tool call, which smuggles a COMMIT past the intended read-only guard and can chain to COPY ... TO PROGRAM for OS code execution. Scored identically to the original CVE-2026-12045 record for consistency, since this is the same vulnerability surviving an incomplete fix rather than a functionally different bug: S:C for the confused-deputy pattern (attacker is not the LLM user; third-party DB content steers the user\u0027s own LLM session into unauthorised writes)."
            }
          ]
        },
        {
          "cvssV4_0": {
            "baseScore": 9.4,
            "baseSeverity": "CRITICAL",
            "vectorString": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H",
            "version": "4.0"
          },
          "format": "CVSS",
          "scenarios": [
            {
              "lang": "en",
              "value": "Same reasoning as the CVSS 3.1 entry and as the original CVE-2026-12045 record: indirect prompt injection, third-party content drives the user\u0027s LLM session, COPY ... TO PROGRAM reaches the DB host. VC:H/VI:H/VA:H + SC:H/SI:H/SA:H since RCE on the DB host compromises both pgAdmin\u0027s authority and the downstream system. UI:P captures that the user merely uses the AI Assistant normally."
            }
          ]
        }
      ],
      "problemTypes": [
        {
          "descriptions": [
            {
              "cweId": "CWE-89",
              "description": "CWE-89 Improper Neutralization of Special Elements used in an SQL Command (\u0027SQL Injection\u0027)",
              "lang": "en",
              "type": "CWE"
            },
            {
              "cweId": "CWE-115",
              "description": "CWE-115 Misinterpretation of Input",
              "lang": "en",
              "type": "CWE"
            }
          ]
        }
      ],
      "providerMetadata": {
        "dateUpdated": "2026-07-31T16:00:19.103Z",
        "orgId": "f86ef6dc-4d3a-42ad-8f28-e6d5547a5007",
        "shortName": "PostgreSQL"
      },
      "references": [
        {
          "tags": [
            "issue-tracking"
          ],
          "url": "https://github.com/pgadmin-org/pgadmin4/issues/10192"
        },
        {
          "tags": [
            "patch"
          ],
          "url": "https://github.com/pgadmin-org/pgadmin4/commit/ef76102bcd1cdb544eb9b4ef18d3382f22b76752"
        },
        {
          "tags": [
            "related"
          ],
          "url": "https://github.com/pgadmin-org/pgadmin4/commit/bf4792444446f0e7ab721d23cbd6bfe6afaa7a8b"
        }
      ],
      "source": {
        "discovery": "EXTERNAL"
      },
      "title": "pgAdmin 4: AI Assistant read-only transaction bypass via sqlparse/PostgreSQL lexer disagreement (incomplete fix for CVE-2026-12045)"
    }
  },
  "cveMetadata": {
    "assignerOrgId": "f86ef6dc-4d3a-42ad-8f28-e6d5547a5007",
    "assignerShortName": "PostgreSQL",
    "cveId": "CVE-2026-17351",
    "datePublished": "2026-07-31T16:00:19.103Z",
    "dateReserved": "2026-07-25T02:52:57.350Z",
    "dateUpdated": "2026-08-01T03:56:14.973Z",
    "state": "PUBLISHED"
  },
  "dataType": "CVE_RECORD",
  "dataVersion": "5.2",
  "vulnerability-lookup:meta": {
    "epss": {
      "cve": "CVE-2026-17351",
      "date": "2026-08-01",
      "epss": "0.00447",
      "percentile": "0.36748"
    },
    "nvd": "{\"cve\":{\"id\":\"CVE-2026-17351\",\"sourceIdentifier\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\",\"published\":\"2026-07-31T16:16:59.807\",\"lastModified\":\"2026-08-01T05:16:55.670\",\"vulnStatus\":\"Received\",\"cveTags\":[],\"descriptions\":[{\"lang\":\"en\",\"value\":\"The fix for CVE-2026-12045 in pgAdmin 4 9.16 required the LLM-supplied query passed to the AI Assistant\u0027s execute_sql_query tool to parse, via sqlparse, as exactly one non-transaction-control statement before running it inside a BEGIN TRANSACTION READ ONLY wrapper. sqlparse\u0027s string-literal lexing can disagree with PostgreSQL\u0027s own parser: under standard_conforming_strings = on (PostgreSQL\u0027s default since 9.1), a backslash immediately before a quote is an ordinary character to PostgreSQL, but sqlparse treats it as escaping the quote. A payload such as SELECT \u0027\\\\\u0027;COMMIT;CREATE TABLE pwn(x int);SELECT 1 --\u0027 therefore parses as a single SELECT to sqlparse\u0027s validator, while PostgreSQL executes it as four statements: the smuggled COMMIT ends the wrapping read-only transaction, and the trailing ROLLBACK becomes a no-op. This reintroduces the same write/RCE bypass CVE-2026-12045 was meant to close, reachable via the same indirect prompt-injection delivery (an attacker plants the payload in any object the AI Assistant may read; the LLM emits it as a tool call).\\n\\nAn initial candidate fix ran the query with psycopg\u0027s execute(..., prepare=True), intending to force PostgreSQL\u0027s own Parse step (extended query protocol) to reject multi-statement text regardless of sqlparse\u0027s classification. This candidate fix does not work as submitted: psycopg3\u0027s PrepareManager silently ignores the prepare argument whenever the connection\u0027s prepare_threshold is None, which is pgAdmin\u0027s default for every server connection (the per-server \\\"Prepare threshold\\\" field is blank unless an administrator explicitly sets it) -- psycopg3 falls back to the simple query protocol, the same multi-statement-capable path the bypass exploits, so the candidate fix closes nothing on any real-world default configuration.\\n\\nThe corrected fix sets conn.prepare_threshold = 0 directly on the dedicated, single-use read-only connection the AI Assistant tool opens, structurally forcing the extended query protocol independent of any server-level configuration. Verified against a live PostgreSQL 18 instance: the payload executes successfully under the prepare_threshold=None (default) behavior, and is rejected with \\\"cannot insert multiple commands into a prepared statement\\\" once prepare_threshold=0 is set on that connection.\\n\\nThis issue affects pgAdmin 4: from 9.13 before 9.17.\"}],\"affected\":[{\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\",\"affectedData\":[{\"vendor\":\"pgadmin.org\",\"product\":\"pgAdmin 4\",\"defaultStatus\":\"unaffected\",\"modules\":[\"AI Assistant\",\"LLM Tools\"],\"programFiles\":[\"https://github.com/pgadmin-org/pgadmin4/blob/master/web/pgadmin/llm/tools/database.py\"],\"repo\":\"https://github.com/pgadmin-org/pgadmin4\",\"versions\":[{\"version\":\"9.13\",\"lessThan\":\"9.17\",\"versionType\":\"custom\",\"status\":\"affected\"}]}]}],\"metrics\":{\"cvssMetricV40\":[{\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\",\"type\":\"Secondary\",\"cvssData\":{\"version\":\"4.0\",\"vectorString\":\"CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X\",\"baseScore\":9.4,\"baseSeverity\":\"CRITICAL\",\"attackVector\":\"NETWORK\",\"attackComplexity\":\"LOW\",\"attackRequirements\":\"NONE\",\"privilegesRequired\":\"LOW\",\"userInteraction\":\"PASSIVE\",\"vulnConfidentialityImpact\":\"HIGH\",\"vulnIntegrityImpact\":\"HIGH\",\"vulnAvailabilityImpact\":\"HIGH\",\"subConfidentialityImpact\":\"HIGH\",\"subIntegrityImpact\":\"HIGH\",\"subAvailabilityImpact\":\"HIGH\",\"exploitMaturity\":\"NOT_DEFINED\",\"confidentialityRequirement\":\"NOT_DEFINED\",\"integrityRequirement\":\"NOT_DEFINED\",\"availabilityRequirement\":\"NOT_DEFINED\",\"modifiedAttackVector\":\"NOT_DEFINED\",\"modifiedAttackComplexity\":\"NOT_DEFINED\",\"modifiedAttackRequirements\":\"NOT_DEFINED\",\"modifiedPrivilegesRequired\":\"NOT_DEFINED\",\"modifiedUserInteraction\":\"NOT_DEFINED\",\"modifiedVulnConfidentialityImpact\":\"NOT_DEFINED\",\"modifiedVulnIntegrityImpact\":\"NOT_DEFINED\",\"modifiedVulnAvailabilityImpact\":\"NOT_DEFINED\",\"modifiedSubConfidentialityImpact\":\"NOT_DEFINED\",\"modifiedSubIntegrityImpact\":\"NOT_DEFINED\",\"modifiedSubAvailabilityImpact\":\"NOT_DEFINED\",\"Safety\":\"NOT_DEFINED\",\"Automatable\":\"NOT_DEFINED\",\"Recovery\":\"NOT_DEFINED\",\"valueDensity\":\"NOT_DEFINED\",\"vulnerabilityResponseEffort\":\"NOT_DEFINED\",\"providerUrgency\":\"NOT_DEFINED\"}}],\"cvssMetricV31\":[{\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\",\"type\":\"Secondary\",\"cvssData\":{\"version\":\"3.1\",\"vectorString\":\"CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H\",\"baseScore\":9.0,\"baseSeverity\":\"CRITICAL\",\"attackVector\":\"NETWORK\",\"attackComplexity\":\"LOW\",\"privilegesRequired\":\"LOW\",\"userInteraction\":\"REQUIRED\",\"scope\":\"CHANGED\",\"confidentialityImpact\":\"HIGH\",\"integrityImpact\":\"HIGH\",\"availabilityImpact\":\"HIGH\"},\"exploitabilityScore\":2.3,\"impactScore\":6.0}],\"ssvcV203\":[{\"source\":\"134c704f-9b21-4f2e-91b3-4a467353bcc0\",\"ssvcData\":{\"timestamp\":\"2026-07-31T00:00:00+00:00\",\"id\":\"CVE-2026-17351\",\"options\":[{\"exploitation\":\"none\"},{\"automatable\":\"no\"},{\"technicalImpact\":\"total\"}],\"role\":\"CISA Coordinator\",\"version\":\"2.0.3\"}}]},\"weaknesses\":[{\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\",\"type\":\"Secondary\",\"description\":[{\"lang\":\"en\",\"value\":\"CWE-89\"},{\"lang\":\"en\",\"value\":\"CWE-115\"}]}],\"references\":[{\"url\":\"https://github.com/pgadmin-org/pgadmin4/commit/bf4792444446f0e7ab721d23cbd6bfe6afaa7a8b\",\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\"},{\"url\":\"https://github.com/pgadmin-org/pgadmin4/commit/ef76102bcd1cdb544eb9b4ef18d3382f22b76752\",\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\"},{\"url\":\"https://github.com/pgadmin-org/pgadmin4/issues/10192\",\"source\":\"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\"}]}}",
    "vulnrichment": {
      "containers": "{\"adp\": [{\"title\": \"CISA ADP Vulnrichment\", \"metrics\": [{\"other\": {\"type\": \"ssvc\", \"content\": {\"id\": \"CVE-2026-17351\", \"role\": \"CISA Coordinator\", \"options\": [{\"Exploitation\": \"none\"}, {\"Automatable\": \"no\"}, {\"Technical Impact\": \"total\"}], \"version\": \"2.0.3\", \"timestamp\": \"2026-07-31T17:00:07.232339Z\"}}}], \"providerMetadata\": {\"orgId\": \"134c704f-9b21-4f2e-91b3-4a467353bcc0\", \"shortName\": \"CISA-ADP\", \"dateUpdated\": \"2026-07-31T17:00:14.067Z\"}}], \"cna\": {\"title\": \"pgAdmin 4: AI Assistant read-only transaction bypass via sqlparse/PostgreSQL lexer disagreement (incomplete fix for CVE-2026-12045)\", \"source\": {\"discovery\": \"EXTERNAL\"}, \"credits\": [{\"lang\": \"en\", \"type\": \"finder\", \"value\": \"Kai Aizen (SnailSploit)\"}, {\"lang\": \"en\", \"type\": \"remediation developer\", \"value\": \"Kundan Sable \u003ckundan.sable@enterprisedb.com\u003e\"}, {\"lang\": \"en\", \"type\": \"remediation developer\", \"value\": \"Ashesh Vashi \u003cashesh.vashi@enterprisedb.com\u003e\"}, {\"lang\": \"en\", \"type\": \"remediation reviewer\", \"value\": \"Ashesh Vashi \u003cashesh.vashi@enterprisedb.com\u003e\"}], \"metrics\": [{\"format\": \"CVSS\", \"cvssV3_1\": {\"scope\": \"CHANGED\", \"version\": \"3.1\", \"baseScore\": 9, \"attackVector\": \"NETWORK\", \"baseSeverity\": \"CRITICAL\", \"vectorString\": \"CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H\", \"integrityImpact\": \"HIGH\", \"userInteraction\": \"REQUIRED\", \"attackComplexity\": \"LOW\", \"availabilityImpact\": \"HIGH\", \"privilegesRequired\": \"LOW\", \"confidentialityImpact\": \"HIGH\"}, \"scenarios\": [{\"lang\": \"en\", \"value\": \"Threat model: identical to CVE-2026-12045 -- indirect prompt injection. An attacker with low-privilege DB write access plants a payload in any object the AI Assistant might read (table row, column value, comment). When a pgAdmin user with a higher-privilege database role (typically superuser or holder of pg_execute_server_program) opens the AI Assistant against that content, the LLM emits the lexer-differential payload as a tool call, which smuggles a COMMIT past the intended read-only guard and can chain to COPY ... TO PROGRAM for OS code execution. Scored identically to the original CVE-2026-12045 record for consistency, since this is the same vulnerability surviving an incomplete fix rather than a functionally different bug: S:C for the confused-deputy pattern (attacker is not the LLM user; third-party DB content steers the user\u0027s own LLM session into unauthorised writes).\"}]}, {\"format\": \"CVSS\", \"cvssV4_0\": {\"version\": \"4.0\", \"baseScore\": 9.4, \"baseSeverity\": \"CRITICAL\", \"vectorString\": \"CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H\"}, \"scenarios\": [{\"lang\": \"en\", \"value\": \"Same reasoning as the CVSS 3.1 entry and as the original CVE-2026-12045 record: indirect prompt injection, third-party content drives the user\u0027s LLM session, COPY ... TO PROGRAM reaches the DB host. VC:H/VI:H/VA:H + SC:H/SI:H/SA:H since RCE on the DB host compromises both pgAdmin\u0027s authority and the downstream system. UI:P captures that the user merely uses the AI Assistant normally.\"}]}], \"affected\": [{\"repo\": \"https://github.com/pgadmin-org/pgadmin4\", \"vendor\": \"pgadmin.org\", \"modules\": [\"AI Assistant\", \"LLM Tools\"], \"product\": \"pgAdmin 4\", \"versions\": [{\"status\": \"affected\", \"version\": \"9.13\", \"lessThan\": \"9.17\", \"versionType\": \"custom\"}], \"programFiles\": [\"https://github.com/pgadmin-org/pgadmin4/blob/master/web/pgadmin/llm/tools/database.py\"], \"defaultStatus\": \"unaffected\"}], \"references\": [{\"url\": \"https://github.com/pgadmin-org/pgadmin4/issues/10192\", \"tags\": [\"issue-tracking\"]}, {\"url\": \"https://github.com/pgadmin-org/pgadmin4/commit/ef76102bcd1cdb544eb9b4ef18d3382f22b76752\", \"tags\": [\"patch\"]}, {\"url\": \"https://github.com/pgadmin-org/pgadmin4/commit/bf4792444446f0e7ab721d23cbd6bfe6afaa7a8b\", \"tags\": [\"related\"]}], \"descriptions\": [{\"lang\": \"en\", \"value\": \"The fix for CVE-2026-12045 in pgAdmin 4 9.16 required the LLM-supplied query passed to the AI Assistant\u0027s execute_sql_query tool to parse, via sqlparse, as exactly one non-transaction-control statement before running it inside a BEGIN TRANSACTION READ ONLY wrapper. sqlparse\u0027s string-literal lexing can disagree with PostgreSQL\u0027s own parser: under standard_conforming_strings = on (PostgreSQL\u0027s default since 9.1), a backslash immediately before a quote is an ordinary character to PostgreSQL, but sqlparse treats it as escaping the quote. A payload such as SELECT \u0027\\\\\u0027;COMMIT;CREATE TABLE pwn(x int);SELECT 1 --\u0027 therefore parses as a single SELECT to sqlparse\u0027s validator, while PostgreSQL executes it as four statements: the smuggled COMMIT ends the wrapping read-only transaction, and the trailing ROLLBACK becomes a no-op. This reintroduces the same write/RCE bypass CVE-2026-12045 was meant to close, reachable via the same indirect prompt-injection delivery (an attacker plants the payload in any object the AI Assistant may read; the LLM emits it as a tool call).\\n\\nAn initial candidate fix ran the query with psycopg\u0027s execute(..., prepare=True), intending to force PostgreSQL\u0027s own Parse step (extended query protocol) to reject multi-statement text regardless of sqlparse\u0027s classification. This candidate fix does not work as submitted: psycopg3\u0027s PrepareManager silently ignores the prepare argument whenever the connection\u0027s prepare_threshold is None, which is pgAdmin\u0027s default for every server connection (the per-server \\\"Prepare threshold\\\" field is blank unless an administrator explicitly sets it) -- psycopg3 falls back to the simple query protocol, the same multi-statement-capable path the bypass exploits, so the candidate fix closes nothing on any real-world default configuration.\\n\\nThe corrected fix sets conn.prepare_threshold = 0 directly on the dedicated, single-use read-only connection the AI Assistant tool opens, structurally forcing the extended query protocol independent of any server-level configuration. Verified against a live PostgreSQL 18 instance: the payload executes successfully under the prepare_threshold=None (default) behavior, and is rejected with \\\"cannot insert multiple commands into a prepared statement\\\" once prepare_threshold=0 is set on that connection.\\n\\nThis issue affects pgAdmin 4: from 9.13 before 9.17.\"}], \"problemTypes\": [{\"descriptions\": [{\"lang\": \"en\", \"type\": \"CWE\", \"cweId\": \"CWE-89\", \"description\": \"CWE-89 Improper Neutralization of Special Elements used in an SQL Command (\u0027SQL Injection\u0027)\"}, {\"lang\": \"en\", \"type\": \"CWE\", \"cweId\": \"CWE-115\", \"description\": \"CWE-115 Misinterpretation of Input\"}]}], \"providerMetadata\": {\"orgId\": \"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\", \"shortName\": \"PostgreSQL\", \"dateUpdated\": \"2026-07-31T16:00:19.103Z\"}}}",
      "cveMetadata": "{\"cveId\": \"CVE-2026-17351\", \"state\": \"PUBLISHED\", \"dateUpdated\": \"2026-08-01T03:56:14.973Z\", \"dateReserved\": \"2026-07-25T02:52:57.350Z\", \"assignerOrgId\": \"f86ef6dc-4d3a-42ad-8f28-e6d5547a5007\", \"datePublished\": \"2026-07-31T16:00:19.103Z\", \"assignerShortName\": \"PostgreSQL\"}",
      "dataType": "CVE_RECORD",
      "dataVersion": "5.2"
    }
  }
}

GHSA-V5GM-27GP-4F7W

Vulnerability from github – Published: 2026-07-31 18:32 – Updated: 2026-07-31 18:32

Details

An initial candidate fix ran the query with psycopg's execute(..., prepare=True), intending to force PostgreSQL's own Parse step (extended query protocol) to reject multi-statement text regardless of sqlparse's classification. This candidate fix does not work as submitted: psycopg3's PrepareManager silently ignores the prepare argument whenever the connection's prepare_threshold is None, which is pgAdmin's default for every server connection (the per-server "Prepare threshold" field is blank unless an administrator explicitly sets it) -- psycopg3 falls back to the simple query protocol, the same multi-statement-capable path the bypass exploits, so the candidate fix closes nothing on any real-world default configuration.

The corrected fix sets conn.prepare_threshold = 0 directly on the dedicated, single-use read-only connection the AI Assistant tool opens, structurally forcing the extended query protocol independent of any server-level configuration. Verified against a live PostgreSQL 18 instance: the payload executes successfully under the prepare_threshold=None (default) behavior, and is rejected with "cannot insert multiple commands into a prepared statement" once prepare_threshold=0 is set on that connection.

This issue affects pgAdmin 4: from 9.13 before 9.17.

Severity

9.0 (Critical)


                  
                    CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H

9.4 (Critical)


                  
                    CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H

Show details on source website

JSON

To clipboard

{
  "affected": [],
  "aliases": [
    "CVE-2026-17351"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-89"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-07-31T16:16:59Z",
    "severity": "CRITICAL"
  },
  "details": "The fix for CVE-2026-12045 in pgAdmin 4 9.16 required the LLM-supplied query passed to the AI Assistant\u0027s execute_sql_query tool to parse, via sqlparse, as exactly one non-transaction-control statement before running it inside a BEGIN TRANSACTION READ ONLY wrapper. sqlparse\u0027s string-literal lexing can disagree with PostgreSQL\u0027s own parser: under standard_conforming_strings = on (PostgreSQL\u0027s default since 9.1), a backslash immediately before a quote is an ordinary character to PostgreSQL, but sqlparse treats it as escaping the quote. A payload such as SELECT \u0027\\\u0027;COMMIT;CREATE TABLE pwn(x int);SELECT 1 --\u0027 therefore parses as a single SELECT to sqlparse\u0027s validator, while PostgreSQL executes it as four statements: the smuggled COMMIT ends the wrapping read-only transaction, and the trailing ROLLBACK becomes a no-op. This reintroduces the same write/RCE bypass CVE-2026-12045 was meant to close, reachable via the same indirect prompt-injection delivery (an attacker plants the payload in any object the AI Assistant may read; the LLM emits it as a tool call).\n\nAn initial candidate fix ran the query with psycopg\u0027s execute(..., prepare=True), intending to force PostgreSQL\u0027s own Parse step (extended query protocol) to reject multi-statement text regardless of sqlparse\u0027s classification. This candidate fix does not work as submitted: psycopg3\u0027s PrepareManager silently ignores the prepare argument whenever the connection\u0027s prepare_threshold is None, which is pgAdmin\u0027s default for every server connection (the per-server \"Prepare threshold\" field is blank unless an administrator explicitly sets it) -- psycopg3 falls back to the simple query protocol, the same multi-statement-capable path the bypass exploits, so the candidate fix closes nothing on any real-world default configuration.\n\nThe corrected fix sets conn.prepare_threshold = 0 directly on the dedicated, single-use read-only connection the AI Assistant tool opens, structurally forcing the extended query protocol independent of any server-level configuration. Verified against a live PostgreSQL 18 instance: the payload executes successfully under the prepare_threshold=None (default) behavior, and is rejected with \"cannot insert multiple commands into a prepared statement\" once prepare_threshold=0 is set on that connection.\n\nThis issue affects pgAdmin 4: from 9.13 before 9.17.",
  "id": "GHSA-v5gm-27gp-4f7w",
  "modified": "2026-07-31T18:32:17Z",
  "published": "2026-07-31T18:32:17Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-17351"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pgadmin-org/pgadmin4/issues/10192"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pgadmin-org/pgadmin4/commit/bf4792444446f0e7ab721d23cbd6bfe6afaa7a8b"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pgadmin-org/pgadmin4/commit/ef76102bcd1cdb544eb9b4ef18d3382f22b76752"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:H/A:H",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:P/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

Sightings

Author	Source	Type	Date	Other

Nomenclature

Seen: The vulnerability was mentioned, discussed, or observed by the user.
Confirmed: The vulnerability has been validated from an analyst's perspective.
Published Proof of Concept: A public proof of concept is available for this vulnerability.
Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
Not confirmed: The user expressed doubt about the validity of the vulnerability.
Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Detection rules are retrieved from Rulezet.

The MITRE ATT&CK techniques below are AI-generated suggestions, inferred from the description of the vulnerability by the CIRCL/vulnerability-attack-technique-classification-roberta-base model, served locally by ML-Gateway. They have not been verified by an analyst and are provided for guidance only.

Action not permitted

CVE-2026-17351 (GCVE-0-2026-17351)

GHSA-V5GM-27GP-4F7W

Tags

Sightings

Nomenclature